fs/f2fs/super.c

   1 /*
   2  * fs/f2fs/super.c
   3  *
   4  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
   5  *             http://www.samsung.com/
   6  *
   7  * This program is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License version 2 as
   9  * published by the Free Software Foundation.
  10  */
  11 #include <linux/module.h>
  12 #include <linux/init.h>
  13 #include <linux/fs.h>
  14 #include <linux/statfs.h>
  15 #include <linux/buffer_head.h>
  16 #include <linux/backing-dev.h>
  17 #include <linux/kthread.h>
  18 #include <linux/parser.h>
  19 #include <linux/mount.h>
  20 #include <linux/seq_file.h>
  21 #include <linux/proc_fs.h>
  22 #include <linux/random.h>
  23 #include <linux/exportfs.h>
  24 #include <linux/blkdev.h>
  25 #include <linux/quotaops.h>
  26 #include <linux/f2fs_fs.h>
  27 #include <linux/sysfs.h>
  28 #include <linux/quota.h>
  29
  30 #include "f2fs.h"
  31 #include "node.h"
  32 #include "segment.h"
  33 #include "xattr.h"
  34 #include "gc.h"
  35 #include "trace.h"
  36
  37 #define CREATE_TRACE_POINTS
  38 #include <trace/events/f2fs.h>
  39
  40 static struct kmem_cache *f2fs_inode_cachep;
  41
  42 #ifdef CONFIG_F2FS_FAULT_INJECTION
  43
  44 char *fault_name[FAULT_MAX] = {
  45         [FAULT_KMALLOC]         = "kmalloc",
  46         [FAULT_KVMALLOC]        = "kvmalloc",
  47         [FAULT_PAGE_ALLOC]      = "page alloc",
  48         [FAULT_PAGE_GET]        = "page get",
  49         [FAULT_ALLOC_BIO]       = "alloc bio",
  50         [FAULT_ALLOC_NID]       = "alloc nid",
  51         [FAULT_ORPHAN]          = "orphan",
  52         [FAULT_BLOCK]           = "no more block",
  53         [FAULT_DIR_DEPTH]       = "too big dir depth",
  54         [FAULT_EVICT_INODE]     = "evict_inode fail",
  55         [FAULT_TRUNCATE]        = "truncate fail",
  56         [FAULT_IO]              = "IO error",
  57         [FAULT_CHECKPOINT]      = "checkpoint error",
  58 };
  59
  60 static void f2fs_build_fault_attr(struct f2fs_sb_info *sbi,
  61                                                 unsigned int rate)
  62 {
  63         struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
  64
  65         if (rate) {
  66                 atomic_set(&ffi->inject_ops, 0);
  67                 ffi->inject_rate = rate;
  68                 ffi->inject_type = (1 << FAULT_MAX) - 1;
  69         } else {
  70                 memset(ffi, 0, sizeof(struct f2fs_fault_info));
  71         }
  72 }
  73 #endif
  74
  75 /* f2fs-wide shrinker description */
  76 static struct shrinker f2fs_shrinker_info = {
  77         .scan_objects = f2fs_shrink_scan,
  78         .count_objects = f2fs_shrink_count,
  79         .seeks = DEFAULT_SEEKS,
  80 };
  81
  82 enum {
  83         Opt_gc_background,
  84         Opt_disable_roll_forward,
  85         Opt_norecovery,
  86         Opt_discard,
  87         Opt_nodiscard,
  88         Opt_noheap,
  89         Opt_heap,
  90         Opt_user_xattr,
  91         Opt_nouser_xattr,
  92         Opt_acl,
  93         Opt_noacl,
  94         Opt_active_logs,
  95         Opt_disable_ext_identify,
  96         Opt_inline_xattr,
  97         Opt_noinline_xattr,
  98         Opt_inline_xattr_size,
  99         Opt_inline_data,
 100         Opt_inline_dentry,
 101         Opt_noinline_dentry,
 102         Opt_flush_merge,
 103         Opt_noflush_merge,
 104         Opt_nobarrier,
 105         Opt_fastboot,
 106         Opt_extent_cache,
 107         Opt_noextent_cache,
 108         Opt_noinline_data,
 109         Opt_data_flush,
 110         Opt_reserve_root,
 111         Opt_resgid,
 112         Opt_resuid,
 113         Opt_mode,
 114         Opt_io_size_bits,
 115         Opt_fault_injection,
 116         Opt_lazytime,
 117         Opt_nolazytime,
 118         Opt_quota,
 119         Opt_noquota,
 120         Opt_usrquota,
 121         Opt_grpquota,
 122         Opt_prjquota,
 123         Opt_usrjquota,
 124         Opt_grpjquota,
 125         Opt_prjjquota,
 126         Opt_offusrjquota,
 127         Opt_offgrpjquota,
 128         Opt_offprjjquota,
 129         Opt_jqfmt_vfsold,
 130         Opt_jqfmt_vfsv0,
 131         Opt_jqfmt_vfsv1,
 132         Opt_whint,
 133         Opt_alloc,
 134         Opt_fsync,
 135         Opt_test_dummy_encryption,
 136         Opt_err,
 137 };
 138
 139 static match_table_t f2fs_tokens = {
 140         {Opt_gc_background, "background_gc=%s"},
 141         {Opt_disable_roll_forward, "disable_roll_forward"},
 142         {Opt_norecovery, "norecovery"},
 143         {Opt_discard, "discard"},
 144         {Opt_nodiscard, "nodiscard"},
 145         {Opt_noheap, "no_heap"},
 146         {Opt_heap, "heap"},
 147         {Opt_user_xattr, "user_xattr"},
 148         {Opt_nouser_xattr, "nouser_xattr"},
 149         {Opt_acl, "acl"},
 150         {Opt_noacl, "noacl"},
 151         {Opt_active_logs, "active_logs=%u"},
 152         {Opt_disable_ext_identify, "disable_ext_identify"},
 153         {Opt_inline_xattr, "inline_xattr"},
 154         {Opt_noinline_xattr, "noinline_xattr"},
 155         {Opt_inline_xattr_size, "inline_xattr_size=%u"},
 156         {Opt_inline_data, "inline_data"},
 157         {Opt_inline_dentry, "inline_dentry"},
 158         {Opt_noinline_dentry, "noinline_dentry"},
 159         {Opt_flush_merge, "flush_merge"},
 160         {Opt_noflush_merge, "noflush_merge"},
 161         {Opt_nobarrier, "nobarrier"},
 162         {Opt_fastboot, "fastboot"},
 163         {Opt_extent_cache, "extent_cache"},
 164         {Opt_noextent_cache, "noextent_cache"},
 165         {Opt_noinline_data, "noinline_data"},
 166         {Opt_data_flush, "data_flush"},
 167         {Opt_reserve_root, "reserve_root=%u"},
 168         {Opt_resgid, "resgid=%u"},
 169         {Opt_resuid, "resuid=%u"},
 170         {Opt_mode, "mode=%s"},
 171         {Opt_io_size_bits, "io_bits=%u"},
 172         {Opt_fault_injection, "fault_injection=%u"},
 173         {Opt_lazytime, "lazytime"},
 174         {Opt_nolazytime, "nolazytime"},
 175         {Opt_quota, "quota"},
 176         {Opt_noquota, "noquota"},
 177         {Opt_usrquota, "usrquota"},
 178         {Opt_grpquota, "grpquota"},
 179         {Opt_prjquota, "prjquota"},
 180         {Opt_usrjquota, "usrjquota=%s"},
 181         {Opt_grpjquota, "grpjquota=%s"},
 182         {Opt_prjjquota, "prjjquota=%s"},
 183         {Opt_offusrjquota, "usrjquota="},
 184         {Opt_offgrpjquota, "grpjquota="},
 185         {Opt_offprjjquota, "prjjquota="},
 186         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
 187         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
 188         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 189         {Opt_whint, "whint_mode=%s"},
 190         {Opt_alloc, "alloc_mode=%s"},
 191         {Opt_fsync, "fsync_mode=%s"},
 192         {Opt_test_dummy_encryption, "test_dummy_encryption"},
 193         {Opt_err, NULL},
 194 };
 195
 196 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
 197 {
 198         struct va_format vaf;
 199         va_list args;
 200
 201         va_start(args, fmt);
 202         vaf.fmt = fmt;
 203         vaf.va = &args;
 204         printk_ratelimited("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
 205         va_end(args);
 206 }
 207
 208 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
 209 {
 210         block_t limit = (sbi->user_block_count << 1) / 1000;
 211
 212         /* limit is 0.2% */
 213         if (test_opt(sbi, RESERVE_ROOT) &&
 214                         F2FS_OPTION(sbi).root_reserved_blocks > limit) {
 215                 F2FS_OPTION(sbi).root_reserved_blocks = limit;
 216                 f2fs_msg(sbi->sb, KERN_INFO,
 217                         "Reduce reserved blocks for root = %u",
 218                         F2FS_OPTION(sbi).root_reserved_blocks);
 219         }
 220         if (!test_opt(sbi, RESERVE_ROOT) &&
 221                 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
 222                                 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
 223                 !gid_eq(F2FS_OPTION(sbi).s_resgid,
 224                                 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
 225                 f2fs_msg(sbi->sb, KERN_INFO,
 226                         "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
 227                                 from_kuid_munged(&init_user_ns,
 228                                         F2FS_OPTION(sbi).s_resuid),
 229                                 from_kgid_munged(&init_user_ns,
 230                                         F2FS_OPTION(sbi).s_resgid));
 231 }
 232
 233 static void init_once(void *foo)
 234 {
 235         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 236
 237         inode_init_once(&fi->vfs_inode);
 238 }
 239
 240 #ifdef CONFIG_QUOTA
 241 static const char * const quotatypes[] = INITQFNAMES;
 242 #define QTYPE2NAME(t) (quotatypes[t])
 243 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
 244                                                         substring_t *args)
 245 {
 246         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 247         char *qname;
 248         int ret = -EINVAL;
 249
 250         if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
 251                 f2fs_msg(sb, KERN_ERR,
 252                         "Cannot change journaled "
 253                         "quota options when quota turned on");
 254                 return -EINVAL;
 255         }
 256         if (f2fs_sb_has_quota_ino(sb)) {
 257                 f2fs_msg(sb, KERN_INFO,
 258                         "QUOTA feature is enabled, so ignore qf_name");
 259                 return 0;
 260         }
 261
 262         qname = match_strdup(args);
 263         if (!qname) {
 264                 f2fs_msg(sb, KERN_ERR,
 265                         "Not enough memory for storing quotafile name");
 266                 return -EINVAL;
 267         }
 268         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 269                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 270                         ret = 0;
 271                 else
 272                         f2fs_msg(sb, KERN_ERR,
 273                                  "%s quota file already specified",
 274                                  QTYPE2NAME(qtype));
 275                 goto errout;
 276         }
 277         if (strchr(qname, '/')) {
 278                 f2fs_msg(sb, KERN_ERR,
 279                         "quotafile must be on filesystem root");
 280                 goto errout;
 281         }
 282         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 283         set_opt(sbi, QUOTA);
 284         return 0;
 285 errout:
 286         kfree(qname);
 287         return ret;
 288 }
 289
 290 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 291 {
 292         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 293
 294         if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
 295                 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options"
 296                         " when quota turned on");
 297                 return -EINVAL;
 298         }
 299         kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 300         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 301         return 0;
 302 }
 303
 304 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 305 {
 306         /*
 307          * We do the test below only for project quotas. 'usrquota' and
 308          * 'grpquota' mount options are allowed even without quota feature
 309          * to support legacy quotas in quota files.
 310          */
 311         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi->sb)) {
 312                 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. "
 313                          "Cannot enable project quota enforcement.");
 314                 return -1;
 315         }
 316         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
 317                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
 318                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
 319                 if (test_opt(sbi, USRQUOTA) &&
 320                                 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
 321                         clear_opt(sbi, USRQUOTA);
 322
 323                 if (test_opt(sbi, GRPQUOTA) &&
 324                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 325                         clear_opt(sbi, GRPQUOTA);
 326
 327                 if (test_opt(sbi, PRJQUOTA) &&
 328                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 329                         clear_opt(sbi, PRJQUOTA);
 330
 331                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 332                                 test_opt(sbi, PRJQUOTA)) {
 333                         f2fs_msg(sbi->sb, KERN_ERR, "old and new quota "
 334                                         "format mixing");
 335                         return -1;
 336                 }
 337
 338                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 339                         f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format "
 340                                         "not specified");
 341                         return -1;
 342                 }
 343         }
 344
 345         if (f2fs_sb_has_quota_ino(sbi->sb) && F2FS_OPTION(sbi).s_jquota_fmt) {
 346                 f2fs_msg(sbi->sb, KERN_INFO,
 347                         "QUOTA feature is enabled, so ignore jquota_fmt");
 348                 F2FS_OPTION(sbi).s_jquota_fmt = 0;
 349         }
 350         if (f2fs_sb_has_quota_ino(sbi->sb) && f2fs_readonly(sbi->sb)) {
 351                 f2fs_msg(sbi->sb, KERN_INFO,
 352                          "Filesystem with quota feature cannot be mounted RDWR "
 353                          "without CONFIG_QUOTA");
 354                 return -1;
 355         }
 356         return 0;
 357 }
 358 #endif
 359
 360 static int parse_options(struct super_block *sb, char *options)
 361 {
 362         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 363         struct request_queue *q;
 364         substring_t args[MAX_OPT_ARGS];
 365         char *p, *name;
 366         int arg = 0;
 367         kuid_t uid;
 368         kgid_t gid;
 369 #ifdef CONFIG_QUOTA
 370         int ret;
 371 #endif
 372
 373         if (!options)
 374                 return 0;
 375
 376         while ((p = strsep(&options, ",")) != NULL) {
 377                 int token;
 378                 if (!*p)
 379                         continue;
 380                 /*
 381                  * Initialize args struct so we know whether arg was
 382                  * found; some options take optional arguments.
 383                  */
 384                 args[0].to = args[0].from = NULL;
 385                 token = match_token(p, f2fs_tokens, args);
 386
 387                 switch (token) {
 388                 case Opt_gc_background:
 389                         name = match_strdup(&args[0]);
 390
 391                         if (!name)
 392                                 return -ENOMEM;
 393                         if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
 394                                 set_opt(sbi, BG_GC);
 395                                 clear_opt(sbi, FORCE_FG_GC);
 396                         } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
 397                                 clear_opt(sbi, BG_GC);
 398                                 clear_opt(sbi, FORCE_FG_GC);
 399                         } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
 400                                 set_opt(sbi, BG_GC);
 401                                 set_opt(sbi, FORCE_FG_GC);
 402                         } else {
 403                                 kfree(name);
 404                                 return -EINVAL;
 405                         }
 406                         kfree(name);
 407                         break;
 408                 case Opt_disable_roll_forward:
 409                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 410                         break;
 411                 case Opt_norecovery:
 412                         /* this option mounts f2fs with ro */
 413                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 414                         if (!f2fs_readonly(sb))
 415                                 return -EINVAL;
 416                         break;
 417                 case Opt_discard:
 418                         q = bdev_get_queue(sb->s_bdev);
 419                         if (blk_queue_discard(q)) {
 420                                 set_opt(sbi, DISCARD);
 421                         } else if (!f2fs_sb_has_blkzoned(sb)) {
 422                                 f2fs_msg(sb, KERN_WARNING,
 423                                         "mounting with \"discard\" option, but "
 424                                         "the device does not support discard");
 425                         }
 426                         break;
 427                 case Opt_nodiscard:
 428                         if (f2fs_sb_has_blkzoned(sb)) {
 429                                 f2fs_msg(sb, KERN_WARNING,
 430                                         "discard is required for zoned block devices");
 431                                 return -EINVAL;
 432                         }
 433                         clear_opt(sbi, DISCARD);
 434                         break;
 435                 case Opt_noheap:
 436                         set_opt(sbi, NOHEAP);
 437                         break;
 438                 case Opt_heap:
 439                         clear_opt(sbi, NOHEAP);
 440                         break;
 441 #ifdef CONFIG_F2FS_FS_XATTR
 442                 case Opt_user_xattr:
 443                         set_opt(sbi, XATTR_USER);
 444                         break;
 445                 case Opt_nouser_xattr:
 446                         clear_opt(sbi, XATTR_USER);
 447                         break;
 448                 case Opt_inline_xattr:
 449                         set_opt(sbi, INLINE_XATTR);
 450                         break;
 451                 case Opt_noinline_xattr:
 452                         clear_opt(sbi, INLINE_XATTR);
 453                         break;
 454                 case Opt_inline_xattr_size:
 455                         if (args->from && match_int(args, &arg))
 456                                 return -EINVAL;
 457                         set_opt(sbi, INLINE_XATTR_SIZE);
 458                         F2FS_OPTION(sbi).inline_xattr_size = arg;
 459                         break;
 460 #else
 461                 case Opt_user_xattr:
 462                         f2fs_msg(sb, KERN_INFO,
 463                                 "user_xattr options not supported");
 464                         break;
 465                 case Opt_nouser_xattr:
 466                         f2fs_msg(sb, KERN_INFO,
 467                                 "nouser_xattr options not supported");
 468                         break;
 469                 case Opt_inline_xattr:
 470                         f2fs_msg(sb, KERN_INFO,
 471                                 "inline_xattr options not supported");
 472                         break;
 473                 case Opt_noinline_xattr:
 474                         f2fs_msg(sb, KERN_INFO,
 475                                 "noinline_xattr options not supported");
 476                         break;
 477 #endif
 478 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 479                 case Opt_acl:
 480                         set_opt(sbi, POSIX_ACL);
 481                         break;
 482                 case Opt_noacl:
 483                         clear_opt(sbi, POSIX_ACL);
 484                         break;
 485 #else
 486                 case Opt_acl:
 487                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
 488                         break;
 489                 case Opt_noacl:
 490                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
 491                         break;
 492 #endif
 493                 case Opt_active_logs:
 494                         if (args->from && match_int(args, &arg))
 495                                 return -EINVAL;
 496                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 497                                 return -EINVAL;
 498                         F2FS_OPTION(sbi).active_logs = arg;
 499                         break;
 500                 case Opt_disable_ext_identify:
 501                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 502                         break;
 503                 case Opt_inline_data:
 504                         set_opt(sbi, INLINE_DATA);
 505                         break;
 506                 case Opt_inline_dentry:
 507                         set_opt(sbi, INLINE_DENTRY);
 508                         break;
 509                 case Opt_noinline_dentry:
 510                         clear_opt(sbi, INLINE_DENTRY);
 511                         break;
 512                 case Opt_flush_merge:
 513                         set_opt(sbi, FLUSH_MERGE);
 514                         break;
 515                 case Opt_noflush_merge:
 516                         clear_opt(sbi, FLUSH_MERGE);
 517                         break;
 518                 case Opt_nobarrier:
 519                         set_opt(sbi, NOBARRIER);
 520                         break;
 521                 case Opt_fastboot:
 522                         set_opt(sbi, FASTBOOT);
 523                         break;
 524                 case Opt_extent_cache:
 525                         set_opt(sbi, EXTENT_CACHE);
 526                         break;
 527                 case Opt_noextent_cache:
 528                         clear_opt(sbi, EXTENT_CACHE);
 529                         break;
 530                 case Opt_noinline_data:
 531                         clear_opt(sbi, INLINE_DATA);
 532                         break;
 533                 case Opt_data_flush:
 534                         set_opt(sbi, DATA_FLUSH);
 535                         break;
 536                 case Opt_reserve_root:
 537                         if (args->from && match_int(args, &arg))
 538                                 return -EINVAL;
 539                         if (test_opt(sbi, RESERVE_ROOT)) {
 540                                 f2fs_msg(sb, KERN_INFO,
 541                                         "Preserve previous reserve_root=%u",
 542                                         F2FS_OPTION(sbi).root_reserved_blocks);
 543                         } else {
 544                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
 545                                 set_opt(sbi, RESERVE_ROOT);
 546                         }
 547                         break;
 548                 case Opt_resuid:
 549                         if (args->from && match_int(args, &arg))
 550                                 return -EINVAL;
 551                         uid = make_kuid(current_user_ns(), arg);
 552                         if (!uid_valid(uid)) {
 553                                 f2fs_msg(sb, KERN_ERR,
 554                                         "Invalid uid value %d", arg);
 555                                 return -EINVAL;
 556                         }
 557                         F2FS_OPTION(sbi).s_resuid = uid;
 558                         break;
 559                 case Opt_resgid:
 560                         if (args->from && match_int(args, &arg))
 561                                 return -EINVAL;
 562                         gid = make_kgid(current_user_ns(), arg);
 563                         if (!gid_valid(gid)) {
 564                                 f2fs_msg(sb, KERN_ERR,
 565                                         "Invalid gid value %d", arg);
 566                                 return -EINVAL;
 567                         }
 568                         F2FS_OPTION(sbi).s_resgid = gid;
 569                         break;
 570                 case Opt_mode:
 571                         name = match_strdup(&args[0]);
 572
 573                         if (!name)
 574                                 return -ENOMEM;
 575                         if (strlen(name) == 8 &&
 576                                         !strncmp(name, "adaptive", 8)) {
 577                                 if (f2fs_sb_has_blkzoned(sb)) {
 578                                         f2fs_msg(sb, KERN_WARNING,
 579                                                  "adaptive mode is not allowed with "
 580                                                  "zoned block device feature");
 581                                         kfree(name);
 582                                         return -EINVAL;
 583                                 }
 584                                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 585                         } else if (strlen(name) == 3 &&
 586                                         !strncmp(name, "lfs", 3)) {
 587                                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 588                         } else {
 589                                 kfree(name);
 590                                 return -EINVAL;
 591                         }
 592                         kfree(name);
 593                         break;
 594                 case Opt_io_size_bits:
 595                         if (args->from && match_int(args, &arg))
 596                                 return -EINVAL;
 597                         if (arg > __ilog2_u32(BIO_MAX_PAGES)) {
 598                                 f2fs_msg(sb, KERN_WARNING,
 599                                         "Not support %d, larger than %d",
 600                                         1 << arg, BIO_MAX_PAGES);
 601                                 return -EINVAL;
 602                         }
 603                         F2FS_OPTION(sbi).write_io_size_bits = arg;
 604                         break;
 605                 case Opt_fault_injection:
 606                         if (args->from && match_int(args, &arg))
 607                                 return -EINVAL;
 608 #ifdef CONFIG_F2FS_FAULT_INJECTION
 609                         f2fs_build_fault_attr(sbi, arg);
 610                         set_opt(sbi, FAULT_INJECTION);
 611 #else
 612                         f2fs_msg(sb, KERN_INFO,
 613                                 "FAULT_INJECTION was not selected");
 614 #endif
 615                         break;
 616                 case Opt_lazytime:
 617                         sb->s_flags |= SB_LAZYTIME;
 618                         break;
 619                 case Opt_nolazytime:
 620                         sb->s_flags &= ~SB_LAZYTIME;
 621                         break;
 622 #ifdef CONFIG_QUOTA
 623                 case Opt_quota:
 624                 case Opt_usrquota:
 625                         set_opt(sbi, USRQUOTA);
 626                         break;
 627                 case Opt_grpquota:
 628                         set_opt(sbi, GRPQUOTA);
 629                         break;
 630                 case Opt_prjquota:
 631                         set_opt(sbi, PRJQUOTA);
 632                         break;
 633                 case Opt_usrjquota:
 634                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 635                         if (ret)
 636                                 return ret;
 637                         break;
 638                 case Opt_grpjquota:
 639                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 640                         if (ret)
 641                                 return ret;
 642                         break;
 643                 case Opt_prjjquota:
 644                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 645                         if (ret)
 646                                 return ret;
 647                         break;
 648                 case Opt_offusrjquota:
 649                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 650                         if (ret)
 651                                 return ret;
 652                         break;
 653                 case Opt_offgrpjquota:
 654                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 655                         if (ret)
 656                                 return ret;
 657                         break;
 658                 case Opt_offprjjquota:
 659                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 660                         if (ret)
 661                                 return ret;
 662                         break;
 663                 case Opt_jqfmt_vfsold:
 664                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 665                         break;
 666                 case Opt_jqfmt_vfsv0:
 667                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 668                         break;
 669                 case Opt_jqfmt_vfsv1:
 670                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 671                         break;
 672                 case Opt_noquota:
 673                         clear_opt(sbi, QUOTA);
 674                         clear_opt(sbi, USRQUOTA);
 675                         clear_opt(sbi, GRPQUOTA);
 676                         clear_opt(sbi, PRJQUOTA);
 677                         break;
 678 #else
 679                 case Opt_quota:
 680                 case Opt_usrquota:
 681                 case Opt_grpquota:
 682                 case Opt_prjquota:
 683                 case Opt_usrjquota:
 684                 case Opt_grpjquota:
 685                 case Opt_prjjquota:
 686                 case Opt_offusrjquota:
 687                 case Opt_offgrpjquota:
 688                 case Opt_offprjjquota:
 689                 case Opt_jqfmt_vfsold:
 690                 case Opt_jqfmt_vfsv0:
 691                 case Opt_jqfmt_vfsv1:
 692                 case Opt_noquota:
 693                         f2fs_msg(sb, KERN_INFO,
 694                                         "quota operations not supported");
 695                         break;
 696 #endif
 697                 case Opt_whint:
 698                         name = match_strdup(&args[0]);
 699                         if (!name)
 700                                 return -ENOMEM;
 701                         if (strlen(name) == 10 &&
 702                                         !strncmp(name, "user-based", 10)) {
 703                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
 704                         } else if (strlen(name) == 3 &&
 705                                         !strncmp(name, "off", 3)) {
 706                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 707                         } else if (strlen(name) == 8 &&
 708                                         !strncmp(name, "fs-based", 8)) {
 709                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
 710                         } else {
 711                                 kfree(name);
 712                                 return -EINVAL;
 713                         }
 714                         kfree(name);
 715                         break;
 716                 case Opt_alloc:
 717                         name = match_strdup(&args[0]);
 718                         if (!name)
 719                                 return -ENOMEM;
 720
 721                         if (strlen(name) == 7 &&
 722                                         !strncmp(name, "default", 7)) {
 723                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
 724                         } else if (strlen(name) == 5 &&
 725                                         !strncmp(name, "reuse", 5)) {
 726                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
 727                         } else {
 728                                 kfree(name);
 729                                 return -EINVAL;
 730                         }
 731                         kfree(name);
 732                         break;
 733                 case Opt_fsync:
 734                         name = match_strdup(&args[0]);
 735                         if (!name)
 736                                 return -ENOMEM;
 737                         if (strlen(name) == 5 &&
 738                                         !strncmp(name, "posix", 5)) {
 739                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
 740                         } else if (strlen(name) == 6 &&
 741                                         !strncmp(name, "strict", 6)) {
 742                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
 743                         } else {
 744                                 kfree(name);
 745                                 return -EINVAL;
 746                         }
 747                         kfree(name);
 748                         break;
 749                 case Opt_test_dummy_encryption:
 750 #ifdef CONFIG_F2FS_FS_ENCRYPTION
 751                         if (!f2fs_sb_has_encrypt(sb)) {
 752                                 f2fs_msg(sb, KERN_ERR, "Encrypt feature is off");
 753                                 return -EINVAL;
 754                         }
 755
 756                         F2FS_OPTION(sbi).test_dummy_encryption = true;
 757                         f2fs_msg(sb, KERN_INFO,
 758                                         "Test dummy encryption mode enabled");
 759 #else
 760                         f2fs_msg(sb, KERN_INFO,
 761                                         "Test dummy encryption mount option ignored");
 762 #endif
 763                         break;
 764                 default:
 765                         f2fs_msg(sb, KERN_ERR,
 766                                 "Unrecognized mount option \"%s\" or missing value",
 767                                 p);
 768                         return -EINVAL;
 769                 }
 770         }
 771 #ifdef CONFIG_QUOTA
 772         if (f2fs_check_quota_options(sbi))
 773                 return -EINVAL;
 774 #endif
 775
 776         if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
 777                 f2fs_msg(sb, KERN_ERR,
 778                                 "Should set mode=lfs with %uKB-sized IO",
 779                                 F2FS_IO_SIZE_KB(sbi));
 780                 return -EINVAL;
 781         }
 782
 783         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
 784                 if (!f2fs_sb_has_extra_attr(sb) ||
 785                         !f2fs_sb_has_flexible_inline_xattr(sb)) {
 786                         f2fs_msg(sb, KERN_ERR,
 787                                         "extra_attr or flexible_inline_xattr "
 788                                         "feature is off");
 789                         return -EINVAL;
 790                 }
 791                 if (!test_opt(sbi, INLINE_XATTR)) {
 792                         f2fs_msg(sb, KERN_ERR,
 793                                         "inline_xattr_size option should be "
 794                                         "set with inline_xattr option");
 795                         return -EINVAL;
 796                 }
 797                 if (!F2FS_OPTION(sbi).inline_xattr_size ||
 798                         F2FS_OPTION(sbi).inline_xattr_size >=
 799                                         DEF_ADDRS_PER_INODE -
 800                                         F2FS_TOTAL_EXTRA_ATTR_SIZE -
 801                                         DEF_INLINE_RESERVED_SIZE -
 802                                         DEF_MIN_INLINE_SIZE) {
 803                         f2fs_msg(sb, KERN_ERR,
 804                                         "inline xattr size is out of range");
 805                         return -EINVAL;
 806                 }
 807         }
 808
 809         /* Not pass down write hints if the number of active logs is lesser
 810          * than NR_CURSEG_TYPE.
 811          */
 812         if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
 813                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 814         return 0;
 815 }
 816
 817 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 818 {
 819         struct f2fs_inode_info *fi;
 820
 821         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 822         if (!fi)
 823                 return NULL;
 824
 825         init_once((void *) fi);
 826
 827         /* Initialize f2fs-specific inode info */
 828         atomic_set(&fi->dirty_pages, 0);
 829         fi->i_current_depth = 1;
 830         init_rwsem(&fi->i_sem);
 831         INIT_LIST_HEAD(&fi->dirty_list);
 832         INIT_LIST_HEAD(&fi->gdirty_list);
 833         INIT_LIST_HEAD(&fi->inmem_ilist);
 834         INIT_LIST_HEAD(&fi->inmem_pages);
 835         mutex_init(&fi->inmem_lock);
 836         init_rwsem(&fi->dio_rwsem[READ]);
 837         init_rwsem(&fi->dio_rwsem[WRITE]);
 838         init_rwsem(&fi->i_mmap_sem);
 839         init_rwsem(&fi->i_xattr_sem);
 840
 841         /* Will be used by directory only */
 842         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 843
 844         return &fi->vfs_inode;
 845 }
 846
 847 static int f2fs_drop_inode(struct inode *inode)
 848 {
 849         int ret;
 850         /*
 851          * This is to avoid a deadlock condition like below.
 852          * writeback_single_inode(inode)
 853          *  - f2fs_write_data_page
 854          *    - f2fs_gc -> iput -> evict
 855          *       - inode_wait_for_writeback(inode)
 856          */
 857         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
 858                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 859                         /* to avoid evict_inode call simultaneously */
 860                         atomic_inc(&inode->i_count);
 861                         spin_unlock(&inode->i_lock);
 862
 863                         /* some remained atomic pages should discarded */
 864                         if (f2fs_is_atomic_file(inode))
 865                                 drop_inmem_pages(inode);
 866
 867                         /* should remain fi->extent_tree for writepage */
 868                         f2fs_destroy_extent_node(inode);
 869
 870                         sb_start_intwrite(inode->i_sb);
 871                         f2fs_i_size_write(inode, 0);
 872
 873                         if (F2FS_HAS_BLOCKS(inode))
 874                                 f2fs_truncate(inode);
 875
 876                         sb_end_intwrite(inode->i_sb);
 877
 878                         spin_lock(&inode->i_lock);
 879                         atomic_dec(&inode->i_count);
 880                 }
 881                 trace_f2fs_drop_inode(inode, 0);
 882                 return 0;
 883         }
 884         ret = generic_drop_inode(inode);
 885         trace_f2fs_drop_inode(inode, ret);
 886         return ret;
 887 }
 888
 889 int f2fs_inode_dirtied(struct inode *inode, bool sync)
 890 {
 891         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 892         int ret = 0;
 893
 894         spin_lock(&sbi->inode_lock[DIRTY_META]);
 895         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 896                 ret = 1;
 897         } else {
 898                 set_inode_flag(inode, FI_DIRTY_INODE);
 899                 stat_inc_dirty_inode(sbi, DIRTY_META);
 900         }
 901         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
 902                 list_add_tail(&F2FS_I(inode)->gdirty_list,
 903                                 &sbi->inode_list[DIRTY_META]);
 904                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
 905         }
 906         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 907         return ret;
 908 }
 909
 910 void f2fs_inode_synced(struct inode *inode)
 911 {
 912         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 913
 914         spin_lock(&sbi->inode_lock[DIRTY_META]);
 915         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 916                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
 917                 return;
 918         }
 919         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
 920                 list_del_init(&F2FS_I(inode)->gdirty_list);
 921                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
 922         }
 923         clear_inode_flag(inode, FI_DIRTY_INODE);
 924         clear_inode_flag(inode, FI_AUTO_RECOVER);
 925         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
 926         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 927 }
 928
 929 /*
 930  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 931  *
 932  * We should call set_dirty_inode to write the dirty inode through write_inode.
 933  */
 934 static void f2fs_dirty_inode(struct inode *inode, int flags)
 935 {
 936         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 937
 938         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 939                         inode->i_ino == F2FS_META_INO(sbi))
 940                 return;
 941
 942         if (flags == I_DIRTY_TIME)
 943                 return;
 944
 945         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
 946                 clear_inode_flag(inode, FI_AUTO_RECOVER);
 947
 948         f2fs_inode_dirtied(inode, false);
 949 }
 950
 951 static void f2fs_i_callback(struct rcu_head *head)
 952 {
 953         struct inode *inode = container_of(head, struct inode, i_rcu);
 954         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
 955 }
 956
 957 static void f2fs_destroy_inode(struct inode *inode)
 958 {
 959         call_rcu(&inode->i_rcu, f2fs_i_callback);
 960 }
 961
 962 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
 963 {
 964         percpu_counter_destroy(&sbi->alloc_valid_block_count);
 965         percpu_counter_destroy(&sbi->total_valid_inode_count);
 966 }
 967
 968 static void destroy_device_list(struct f2fs_sb_info *sbi)
 969 {
 970         int i;
 971
 972         for (i = 0; i < sbi->s_ndevs; i++) {
 973                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
 974 #ifdef CONFIG_BLK_DEV_ZONED
 975                 kfree(FDEV(i).blkz_type);
 976 #endif
 977         }
 978         kfree(sbi->devs);
 979 }
 980
 981 static void f2fs_put_super(struct super_block *sb)
 982 {
 983         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 984         int i;
 985         bool dropped;
 986
 987         f2fs_quota_off_umount(sb);
 988
 989         /* prevent remaining shrinker jobs */
 990         mutex_lock(&sbi->umount_mutex);
 991
 992         /*
 993          * We don't need to do checkpoint when superblock is clean.
 994          * But, the previous checkpoint was not done by umount, it needs to do
 995          * clean checkpoint again.
 996          */
 997         if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
 998                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
 999                 struct cp_control cpc = {
1000                         .reason = CP_UMOUNT,
1001                 };
1002                 write_checkpoint(sbi, &cpc);
1003         }
1004
1005         /* be sure to wait for any on-going discard commands */
1006         dropped = f2fs_wait_discard_bios(sbi);
1007
1008         if (f2fs_discard_en(sbi) && !sbi->discard_blks && !dropped) {
1009                 struct cp_control cpc = {
1010                         .reason = CP_UMOUNT | CP_TRIMMED,
1011                 };
1012                 write_checkpoint(sbi, &cpc);
1013         }
1014
1015         /* write_checkpoint can update stat informaion */
1016         f2fs_destroy_stats(sbi);
1017
1018         /*
1019          * normally superblock is clean, so we need to release this.
1020          * In addition, EIO will skip do checkpoint, we need this as well.
1021          */
1022         release_ino_entry(sbi, true);
1023
1024         f2fs_leave_shrinker(sbi);
1025         mutex_unlock(&sbi->umount_mutex);
1026
1027         /* our cp_error case, we can wait for any writeback page */
1028         f2fs_flush_merged_writes(sbi);
1029
1030         iput(sbi->node_inode);
1031         iput(sbi->meta_inode);
1032
1033         /* destroy f2fs internal modules */
1034         destroy_node_manager(sbi);
1035         destroy_segment_manager(sbi);
1036
1037         kfree(sbi->ckpt);
1038
1039         f2fs_unregister_sysfs(sbi);
1040
1041         sb->s_fs_info = NULL;
1042         if (sbi->s_chksum_driver)
1043                 crypto_free_shash(sbi->s_chksum_driver);
1044         kfree(sbi->raw_super);
1045
1046         destroy_device_list(sbi);
1047         mempool_destroy(sbi->write_io_dummy);
1048 #ifdef CONFIG_QUOTA
1049         for (i = 0; i < MAXQUOTAS; i++)
1050                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1051 #endif
1052         destroy_percpu_info(sbi);
1053         for (i = 0; i < NR_PAGE_TYPE; i++)
1054                 kfree(sbi->write_io[i]);
1055         kfree(sbi);
1056 }
1057
1058 int f2fs_sync_fs(struct super_block *sb, int sync)
1059 {
1060         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1061         int err = 0;
1062
1063         if (unlikely(f2fs_cp_error(sbi)))
1064                 return 0;
1065
1066         trace_f2fs_sync_fs(sb, sync);
1067
1068         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1069                 return -EAGAIN;
1070
1071         if (sync) {
1072                 struct cp_control cpc;
1073
1074                 cpc.reason = __get_cp_reason(sbi);
1075
1076                 mutex_lock(&sbi->gc_mutex);
1077                 err = write_checkpoint(sbi, &cpc);
1078                 mutex_unlock(&sbi->gc_mutex);
1079         }
1080         f2fs_trace_ios(NULL, 1);
1081
1082         return err;
1083 }
1084
1085 static int f2fs_freeze(struct super_block *sb)
1086 {
1087         if (f2fs_readonly(sb))
1088                 return 0;
1089
1090         /* IO error happened before */
1091         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1092                 return -EIO;
1093
1094         /* must be clean, since sync_filesystem() was already called */
1095         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1096                 return -EINVAL;
1097         return 0;
1098 }
1099
1100 static int f2fs_unfreeze(struct super_block *sb)
1101 {
1102         return 0;
1103 }
1104
1105 #ifdef CONFIG_QUOTA
1106 static int f2fs_statfs_project(struct super_block *sb,
1107                                 kprojid_t projid, struct kstatfs *buf)
1108 {
1109         struct kqid qid;
1110         struct dquot *dquot;
1111         u64 limit;
1112         u64 curblock;
1113
1114         qid = make_kqid_projid(projid);
1115         dquot = dqget(sb, qid);
1116         if (IS_ERR(dquot))
1117                 return PTR_ERR(dquot);
1118         spin_lock(&dq_data_lock);
1119
1120         limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1121                  dquot->dq_dqb.dqb_bsoftlimit :
1122                  dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1123         if (limit && buf->f_blocks > limit) {
1124                 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1125                 buf->f_blocks = limit;
1126                 buf->f_bfree = buf->f_bavail =
1127                         (buf->f_blocks > curblock) ?
1128                          (buf->f_blocks - curblock) : 0;
1129         }
1130
1131         limit = dquot->dq_dqb.dqb_isoftlimit ?
1132                 dquot->dq_dqb.dqb_isoftlimit :
1133                 dquot->dq_dqb.dqb_ihardlimit;
1134         if (limit && buf->f_files > limit) {
1135                 buf->f_files = limit;
1136                 buf->f_ffree =
1137                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1138                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1139         }
1140
1141         spin_unlock(&dq_data_lock);
1142         dqput(dquot);
1143         return 0;
1144 }
1145 #endif
1146
1147 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1148 {
1149         struct super_block *sb = dentry->d_sb;
1150         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1151         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1152         block_t total_count, user_block_count, start_count;
1153         u64 avail_node_count;
1154
1155         total_count = le64_to_cpu(sbi->raw_super->block_count);
1156         user_block_count = sbi->user_block_count;
1157         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1158         buf->f_type = F2FS_SUPER_MAGIC;
1159         buf->f_bsize = sbi->blocksize;
1160
1161         buf->f_blocks = total_count - start_count;
1162         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1163                                                 sbi->current_reserved_blocks;
1164         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1165                 buf->f_bavail = buf->f_bfree -
1166                                 F2FS_OPTION(sbi).root_reserved_blocks;
1167         else
1168                 buf->f_bavail = 0;
1169
1170         avail_node_count = sbi->total_node_count - sbi->nquota_files -
1171                                                 F2FS_RESERVED_NODE_NUM;
1172
1173         if (avail_node_count > user_block_count) {
1174                 buf->f_files = user_block_count;
1175                 buf->f_ffree = buf->f_bavail;
1176         } else {
1177                 buf->f_files = avail_node_count;
1178                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1179                                         buf->f_bavail);
1180         }
1181
1182         buf->f_namelen = F2FS_NAME_LEN;
1183         buf->f_fsid.val[0] = (u32)id;
1184         buf->f_fsid.val[1] = (u32)(id >> 32);
1185
1186 #ifdef CONFIG_QUOTA
1187         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1188                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1189                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1190         }
1191 #endif
1192         return 0;
1193 }
1194
1195 static inline void f2fs_show_quota_options(struct seq_file *seq,
1196                                            struct super_block *sb)
1197 {
1198 #ifdef CONFIG_QUOTA
1199         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1200
1201         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1202                 char *fmtname = "";
1203
1204                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1205                 case QFMT_VFS_OLD:
1206                         fmtname = "vfsold";
1207                         break;
1208                 case QFMT_VFS_V0:
1209                         fmtname = "vfsv0";
1210                         break;
1211                 case QFMT_VFS_V1:
1212                         fmtname = "vfsv1";
1213                         break;
1214                 }
1215                 seq_printf(seq, ",jqfmt=%s", fmtname);
1216         }
1217
1218         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1219                 seq_show_option(seq, "usrjquota",
1220                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1221
1222         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1223                 seq_show_option(seq, "grpjquota",
1224                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1225
1226         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1227                 seq_show_option(seq, "prjjquota",
1228                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1229 #endif
1230 }
1231
1232 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1233 {
1234         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1235
1236         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1237                 if (test_opt(sbi, FORCE_FG_GC))
1238                         seq_printf(seq, ",background_gc=%s", "sync");
1239                 else
1240                         seq_printf(seq, ",background_gc=%s", "on");
1241         } else {
1242                 seq_printf(seq, ",background_gc=%s", "off");
1243         }
1244         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1245                 seq_puts(seq, ",disable_roll_forward");
1246         if (test_opt(sbi, DISCARD))
1247                 seq_puts(seq, ",discard");
1248         if (test_opt(sbi, NOHEAP))
1249                 seq_puts(seq, ",no_heap");
1250         else
1251                 seq_puts(seq, ",heap");
1252 #ifdef CONFIG_F2FS_FS_XATTR
1253         if (test_opt(sbi, XATTR_USER))
1254                 seq_puts(seq, ",user_xattr");
1255         else
1256                 seq_puts(seq, ",nouser_xattr");
1257         if (test_opt(sbi, INLINE_XATTR))
1258                 seq_puts(seq, ",inline_xattr");
1259         else
1260                 seq_puts(seq, ",noinline_xattr");
1261         if (test_opt(sbi, INLINE_XATTR_SIZE))
1262                 seq_printf(seq, ",inline_xattr_size=%u",
1263                                         F2FS_OPTION(sbi).inline_xattr_size);
1264 #endif
1265 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1266         if (test_opt(sbi, POSIX_ACL))
1267                 seq_puts(seq, ",acl");
1268         else
1269                 seq_puts(seq, ",noacl");
1270 #endif
1271         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1272                 seq_puts(seq, ",disable_ext_identify");
1273         if (test_opt(sbi, INLINE_DATA))
1274                 seq_puts(seq, ",inline_data");
1275         else
1276                 seq_puts(seq, ",noinline_data");
1277         if (test_opt(sbi, INLINE_DENTRY))
1278                 seq_puts(seq, ",inline_dentry");
1279         else
1280                 seq_puts(seq, ",noinline_dentry");
1281         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1282                 seq_puts(seq, ",flush_merge");
1283         if (test_opt(sbi, NOBARRIER))
1284                 seq_puts(seq, ",nobarrier");
1285         if (test_opt(sbi, FASTBOOT))
1286                 seq_puts(seq, ",fastboot");
1287         if (test_opt(sbi, EXTENT_CACHE))
1288                 seq_puts(seq, ",extent_cache");
1289         else
1290                 seq_puts(seq, ",noextent_cache");
1291         if (test_opt(sbi, DATA_FLUSH))
1292                 seq_puts(seq, ",data_flush");
1293
1294         seq_puts(seq, ",mode=");
1295         if (test_opt(sbi, ADAPTIVE))
1296                 seq_puts(seq, "adaptive");
1297         else if (test_opt(sbi, LFS))
1298                 seq_puts(seq, "lfs");
1299         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1300         if (test_opt(sbi, RESERVE_ROOT))
1301                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1302                                 F2FS_OPTION(sbi).root_reserved_blocks,
1303                                 from_kuid_munged(&init_user_ns,
1304                                         F2FS_OPTION(sbi).s_resuid),
1305                                 from_kgid_munged(&init_user_ns,
1306                                         F2FS_OPTION(sbi).s_resgid));
1307         if (F2FS_IO_SIZE_BITS(sbi))
1308                 seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi));
1309 #ifdef CONFIG_F2FS_FAULT_INJECTION
1310         if (test_opt(sbi, FAULT_INJECTION))
1311                 seq_printf(seq, ",fault_injection=%u",
1312                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1313 #endif
1314 #ifdef CONFIG_QUOTA
1315         if (test_opt(sbi, QUOTA))
1316                 seq_puts(seq, ",quota");
1317         if (test_opt(sbi, USRQUOTA))
1318                 seq_puts(seq, ",usrquota");
1319         if (test_opt(sbi, GRPQUOTA))
1320                 seq_puts(seq, ",grpquota");
1321         if (test_opt(sbi, PRJQUOTA))
1322                 seq_puts(seq, ",prjquota");
1323 #endif
1324         f2fs_show_quota_options(seq, sbi->sb);
1325         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1326                 seq_printf(seq, ",whint_mode=%s", "user-based");
1327         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1328                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1329 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1330         if (F2FS_OPTION(sbi).test_dummy_encryption)
1331                 seq_puts(seq, ",test_dummy_encryption");
1332 #endif
1333
1334         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1335                 seq_printf(seq, ",alloc_mode=%s", "default");
1336         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1337                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1338
1339         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1340                 seq_printf(seq, ",fsync_mode=%s", "posix");
1341         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1342                 seq_printf(seq, ",fsync_mode=%s", "strict");
1343         return 0;
1344 }
1345
1346 static void default_options(struct f2fs_sb_info *sbi)
1347 {
1348         /* init some FS parameters */
1349         F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1350         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1351         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1352         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1353         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1354         F2FS_OPTION(sbi).test_dummy_encryption = false;
1355         sbi->readdir_ra = 1;
1356
1357         set_opt(sbi, BG_GC);
1358         set_opt(sbi, INLINE_XATTR);
1359         set_opt(sbi, INLINE_DATA);
1360         set_opt(sbi, INLINE_DENTRY);
1361         set_opt(sbi, EXTENT_CACHE);
1362         set_opt(sbi, NOHEAP);
1363         sbi->sb->s_flags |= SB_LAZYTIME;
1364         set_opt(sbi, FLUSH_MERGE);
1365         if (f2fs_sb_has_blkzoned(sbi->sb)) {
1366                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1367                 set_opt(sbi, DISCARD);
1368         } else {
1369                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1370         }
1371
1372 #ifdef CONFIG_F2FS_FS_XATTR
1373         set_opt(sbi, XATTR_USER);
1374 #endif
1375 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1376         set_opt(sbi, POSIX_ACL);
1377 #endif
1378
1379 #ifdef CONFIG_F2FS_FAULT_INJECTION
1380         f2fs_build_fault_attr(sbi, 0);
1381 #endif
1382 }
1383
1384 #ifdef CONFIG_QUOTA
1385 static int f2fs_enable_quotas(struct super_block *sb);
1386 #endif
1387 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1388 {
1389         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1390         struct f2fs_mount_info org_mount_opt;
1391         unsigned long old_sb_flags;
1392         int err;
1393         bool need_restart_gc = false;
1394         bool need_stop_gc = false;
1395         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1396 #ifdef CONFIG_QUOTA
1397         int i, j;
1398 #endif
1399
1400         /*
1401          * Save the old mount options in case we
1402          * need to restore them.
1403          */
1404         org_mount_opt = sbi->mount_opt;
1405         old_sb_flags = sb->s_flags;
1406
1407 #ifdef CONFIG_QUOTA
1408         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1409         for (i = 0; i < MAXQUOTAS; i++) {
1410                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1411                         org_mount_opt.s_qf_names[i] =
1412                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1413                                 GFP_KERNEL);
1414                         if (!org_mount_opt.s_qf_names[i]) {
1415                                 for (j = 0; j < i; j++)
1416                                         kfree(org_mount_opt.s_qf_names[j]);
1417                                 return -ENOMEM;
1418                         }
1419                 } else {
1420                         org_mount_opt.s_qf_names[i] = NULL;
1421                 }
1422         }
1423 #endif
1424
1425         /* recover superblocks we couldn't write due to previous RO mount */
1426         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1427                 err = f2fs_commit_super(sbi, false);
1428                 f2fs_msg(sb, KERN_INFO,
1429                         "Try to recover all the superblocks, ret: %d", err);
1430                 if (!err)
1431                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1432         }
1433
1434         default_options(sbi);
1435
1436         /* parse mount options */
1437         err = parse_options(sb, data);
1438         if (err)
1439                 goto restore_opts;
1440
1441         /*
1442          * Previous and new state of filesystem is RO,
1443          * so skip checking GC and FLUSH_MERGE conditions.
1444          */
1445         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1446                 goto skip;
1447
1448 #ifdef CONFIG_QUOTA
1449         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1450                 err = dquot_suspend(sb, -1);
1451                 if (err < 0)
1452                         goto restore_opts;
1453         } else if (f2fs_readonly(sb) && !(*flags & MS_RDONLY)) {
1454                 /* dquot_resume needs RW */
1455                 sb->s_flags &= ~SB_RDONLY;
1456                 if (sb_any_quota_suspended(sb)) {
1457                         dquot_resume(sb, -1);
1458                 } else if (f2fs_sb_has_quota_ino(sb)) {
1459                         err = f2fs_enable_quotas(sb);
1460                         if (err)
1461                                 goto restore_opts;
1462                 }
1463         }
1464 #endif
1465         /* disallow enable/disable extent_cache dynamically */
1466         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1467                 err = -EINVAL;
1468                 f2fs_msg(sbi->sb, KERN_WARNING,
1469                                 "switch extent_cache option is not allowed");
1470                 goto restore_opts;
1471         }
1472
1473         /*
1474          * We stop the GC thread if FS is mounted as RO
1475          * or if background_gc = off is passed in mount
1476          * option. Also sync the filesystem.
1477          */
1478         if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1479                 if (sbi->gc_thread) {
1480                         stop_gc_thread(sbi);
1481                         need_restart_gc = true;
1482                 }
1483         } else if (!sbi->gc_thread) {
1484                 err = start_gc_thread(sbi);
1485                 if (err)
1486                         goto restore_opts;
1487                 need_stop_gc = true;
1488         }
1489
1490         if (*flags & SB_RDONLY ||
1491                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1492                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1493                 sync_inodes_sb(sb);
1494
1495                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1496                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1497                 f2fs_sync_fs(sb, 1);
1498                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1499         }
1500
1501         /*
1502          * We stop issue flush thread if FS is mounted as RO
1503          * or if flush_merge is not passed in mount option.
1504          */
1505         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1506                 clear_opt(sbi, FLUSH_MERGE);
1507                 destroy_flush_cmd_control(sbi, false);
1508         } else {
1509                 err = create_flush_cmd_control(sbi);
1510                 if (err)
1511                         goto restore_gc;
1512         }
1513 skip:
1514 #ifdef CONFIG_QUOTA
1515         /* Release old quota file names */
1516         for (i = 0; i < MAXQUOTAS; i++)
1517                 kfree(org_mount_opt.s_qf_names[i]);
1518 #endif
1519         /* Update the POSIXACL Flag */
1520         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1521                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1522
1523         limit_reserve_root(sbi);
1524         return 0;
1525 restore_gc:
1526         if (need_restart_gc) {
1527                 if (start_gc_thread(sbi))
1528                         f2fs_msg(sbi->sb, KERN_WARNING,
1529                                 "background gc thread has stopped");
1530         } else if (need_stop_gc) {
1531                 stop_gc_thread(sbi);
1532         }
1533 restore_opts:
1534 #ifdef CONFIG_QUOTA
1535         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1536         for (i = 0; i < MAXQUOTAS; i++) {
1537                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1538                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1539         }
1540 #endif
1541         sbi->mount_opt = org_mount_opt;
1542         sb->s_flags = old_sb_flags;
1543         return err;
1544 }
1545
1546 #ifdef CONFIG_QUOTA
1547 /* Read data from quotafile */
1548 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1549                                size_t len, loff_t off)
1550 {
1551         struct inode *inode = sb_dqopt(sb)->files[type];
1552         struct address_space *mapping = inode->i_mapping;
1553         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1554         int offset = off & (sb->s_blocksize - 1);
1555         int tocopy;
1556         size_t toread;
1557         loff_t i_size = i_size_read(inode);
1558         struct page *page;
1559         char *kaddr;
1560
1561         if (off > i_size)
1562                 return 0;
1563
1564         if (off + len > i_size)
1565                 len = i_size - off;
1566         toread = len;
1567         while (toread > 0) {
1568                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1569 repeat:
1570                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1571                 if (IS_ERR(page)) {
1572                         if (PTR_ERR(page) == -ENOMEM) {
1573                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1574                                 goto repeat;
1575                         }
1576                         return PTR_ERR(page);
1577                 }
1578
1579                 lock_page(page);
1580
1581                 if (unlikely(page->mapping != mapping)) {
1582                         f2fs_put_page(page, 1);
1583                         goto repeat;
1584                 }
1585                 if (unlikely(!PageUptodate(page))) {
1586                         f2fs_put_page(page, 1);
1587                         return -EIO;
1588                 }
1589
1590                 kaddr = kmap_atomic(page);
1591                 memcpy(data, kaddr + offset, tocopy);
1592                 kunmap_atomic(kaddr);
1593                 f2fs_put_page(page, 1);
1594
1595                 offset = 0;
1596                 toread -= tocopy;
1597                 data += tocopy;
1598                 blkidx++;
1599         }
1600         return len;
1601 }
1602
1603 /* Write to quotafile */
1604 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1605                                 const char *data, size_t len, loff_t off)
1606 {
1607         struct inode *inode = sb_dqopt(sb)->files[type];
1608         struct address_space *mapping = inode->i_mapping;
1609         const struct address_space_operations *a_ops = mapping->a_ops;
1610         int offset = off & (sb->s_blocksize - 1);
1611         size_t towrite = len;
1612         struct page *page;
1613         char *kaddr;
1614         int err = 0;
1615         int tocopy;
1616
1617         while (towrite > 0) {
1618                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1619                                                                 towrite);
1620 retry:
1621                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1622                                                         &page, NULL);
1623                 if (unlikely(err)) {
1624                         if (err == -ENOMEM) {
1625                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1626                                 goto retry;
1627                         }
1628                         break;
1629                 }
1630
1631                 kaddr = kmap_atomic(page);
1632                 memcpy(kaddr + offset, data, tocopy);
1633                 kunmap_atomic(kaddr);
1634                 flush_dcache_page(page);
1635
1636                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1637                                                 page, NULL);
1638                 offset = 0;
1639                 towrite -= tocopy;
1640                 off += tocopy;
1641                 data += tocopy;
1642                 cond_resched();
1643         }
1644
1645         if (len == towrite)
1646                 return err;
1647         inode->i_mtime = inode->i_ctime = current_time(inode);
1648         f2fs_mark_inode_dirty_sync(inode, false);
1649         return len - towrite;
1650 }
1651
1652 static struct dquot **f2fs_get_dquots(struct inode *inode)
1653 {
1654         return F2FS_I(inode)->i_dquot;
1655 }
1656
1657 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1658 {
1659         return &F2FS_I(inode)->i_reserved_quota;
1660 }
1661
1662 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1663 {
1664         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1665                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
1666 }
1667
1668 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1669 {
1670         int enabled = 0;
1671         int i, err;
1672
1673         if (f2fs_sb_has_quota_ino(sbi->sb) && rdonly) {
1674                 err = f2fs_enable_quotas(sbi->sb);
1675                 if (err) {
1676                         f2fs_msg(sbi->sb, KERN_ERR,
1677                                         "Cannot turn on quota_ino: %d", err);
1678                         return 0;
1679                 }
1680                 return 1;
1681         }
1682
1683         for (i = 0; i < MAXQUOTAS; i++) {
1684                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1685                         err = f2fs_quota_on_mount(sbi, i);
1686                         if (!err) {
1687                                 enabled = 1;
1688                                 continue;
1689                         }
1690                         f2fs_msg(sbi->sb, KERN_ERR,
1691                                 "Cannot turn on quotas: %d on %d", err, i);
1692                 }
1693         }
1694         return enabled;
1695 }
1696
1697 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1698                              unsigned int flags)
1699 {
1700         struct inode *qf_inode;
1701         unsigned long qf_inum;
1702         int err;
1703
1704         BUG_ON(!f2fs_sb_has_quota_ino(sb));
1705
1706         qf_inum = f2fs_qf_ino(sb, type);
1707         if (!qf_inum)
1708                 return -EPERM;
1709
1710         qf_inode = f2fs_iget(sb, qf_inum);
1711         if (IS_ERR(qf_inode)) {
1712                 f2fs_msg(sb, KERN_ERR,
1713                         "Bad quota inode %u:%lu", type, qf_inum);
1714                 return PTR_ERR(qf_inode);
1715         }
1716
1717         /* Don't account quota for quota files to avoid recursion */
1718         qf_inode->i_flags |= S_NOQUOTA;
1719         err = dquot_enable(qf_inode, type, format_id, flags);
1720         iput(qf_inode);
1721         return err;
1722 }
1723
1724 static int f2fs_enable_quotas(struct super_block *sb)
1725 {
1726         int type, err = 0;
1727         unsigned long qf_inum;
1728         bool quota_mopt[MAXQUOTAS] = {
1729                 test_opt(F2FS_SB(sb), USRQUOTA),
1730                 test_opt(F2FS_SB(sb), GRPQUOTA),
1731                 test_opt(F2FS_SB(sb), PRJQUOTA),
1732         };
1733
1734         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
1735         for (type = 0; type < MAXQUOTAS; type++) {
1736                 qf_inum = f2fs_qf_ino(sb, type);
1737                 if (qf_inum) {
1738                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1739                                 DQUOT_USAGE_ENABLED |
1740                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1741                         if (err) {
1742                                 f2fs_msg(sb, KERN_ERR,
1743                                         "Failed to enable quota tracking "
1744                                         "(type=%d, err=%d). Please run "
1745                                         "fsck to fix.", type, err);
1746                                 for (type--; type >= 0; type--)
1747                                         dquot_quota_off(sb, type);
1748                                 return err;
1749                         }
1750                 }
1751         }
1752         return 0;
1753 }
1754
1755 static int f2fs_quota_sync(struct super_block *sb, int type)
1756 {
1757         struct quota_info *dqopt = sb_dqopt(sb);
1758         int cnt;
1759         int ret;
1760
1761         ret = dquot_writeback_dquots(sb, type);
1762         if (ret)
1763                 return ret;
1764
1765         /*
1766          * Now when everything is written we can discard the pagecache so
1767          * that userspace sees the changes.
1768          */
1769         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1770                 if (type != -1 && cnt != type)
1771                         continue;
1772                 if (!sb_has_quota_active(sb, cnt))
1773                         continue;
1774
1775                 ret = filemap_write_and_wait(dqopt->files[cnt]->i_mapping);
1776                 if (ret)
1777                         return ret;
1778
1779                 inode_lock(dqopt->files[cnt]);
1780                 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1781                 inode_unlock(dqopt->files[cnt]);
1782         }
1783         return 0;
1784 }
1785
1786 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1787                                                         const struct path *path)
1788 {
1789         struct inode *inode;
1790         int err;
1791
1792         err = f2fs_quota_sync(sb, type);
1793         if (err)
1794                 return err;
1795
1796         err = dquot_quota_on(sb, type, format_id, path);
1797         if (err)
1798                 return err;
1799
1800         inode = d_inode(path->dentry);
1801
1802         inode_lock(inode);
1803         F2FS_I(inode)->i_flags |= FS_NOATIME_FL | FS_IMMUTABLE_FL;
1804         inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1805                                         S_NOATIME | S_IMMUTABLE);
1806         inode_unlock(inode);
1807         f2fs_mark_inode_dirty_sync(inode, false);
1808
1809         return 0;
1810 }
1811
1812 static int f2fs_quota_off(struct super_block *sb, int type)
1813 {
1814         struct inode *inode = sb_dqopt(sb)->files[type];
1815         int err;
1816
1817         if (!inode || !igrab(inode))
1818                 return dquot_quota_off(sb, type);
1819
1820         f2fs_quota_sync(sb, type);
1821
1822         err = dquot_quota_off(sb, type);
1823         if (err || f2fs_sb_has_quota_ino(sb))
1824                 goto out_put;
1825
1826         inode_lock(inode);
1827         F2FS_I(inode)->i_flags &= ~(FS_NOATIME_FL | FS_IMMUTABLE_FL);
1828         inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1829         inode_unlock(inode);
1830         f2fs_mark_inode_dirty_sync(inode, false);
1831 out_put:
1832         iput(inode);
1833         return err;
1834 }
1835
1836 void f2fs_quota_off_umount(struct super_block *sb)
1837 {
1838         int type;
1839
1840         for (type = 0; type < MAXQUOTAS; type++)
1841                 f2fs_quota_off(sb, type);
1842 }
1843
1844 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
1845 {
1846         *projid = F2FS_I(inode)->i_projid;
1847         return 0;
1848 }
1849
1850 static const struct dquot_operations f2fs_quota_operations = {
1851         .get_reserved_space = f2fs_get_reserved_space,
1852         .write_dquot    = dquot_commit,
1853         .acquire_dquot  = dquot_acquire,
1854         .release_dquot  = dquot_release,
1855         .mark_dirty     = dquot_mark_dquot_dirty,
1856         .write_info     = dquot_commit_info,
1857         .alloc_dquot    = dquot_alloc,
1858         .destroy_dquot  = dquot_destroy,
1859         .get_projid     = f2fs_get_projid,
1860         .get_next_id    = dquot_get_next_id,
1861 };
1862
1863 static const struct quotactl_ops f2fs_quotactl_ops = {
1864         .quota_on       = f2fs_quota_on,
1865         .quota_off      = f2fs_quota_off,
1866         .quota_sync     = f2fs_quota_sync,
1867         .get_state      = dquot_get_state,
1868         .set_info       = dquot_set_dqinfo,
1869         .get_dqblk      = dquot_get_dqblk,
1870         .set_dqblk      = dquot_set_dqblk,
1871         .get_nextdqblk  = dquot_get_next_dqblk,
1872 };
1873 #else
1874 void f2fs_quota_off_umount(struct super_block *sb)
1875 {
1876 }
1877 #endif
1878
1879 static const struct super_operations f2fs_sops = {
1880         .alloc_inode    = f2fs_alloc_inode,
1881         .drop_inode     = f2fs_drop_inode,
1882         .destroy_inode  = f2fs_destroy_inode,
1883         .write_inode    = f2fs_write_inode,
1884         .dirty_inode    = f2fs_dirty_inode,
1885         .show_options   = f2fs_show_options,
1886 #ifdef CONFIG_QUOTA
1887         .quota_read     = f2fs_quota_read,
1888         .quota_write    = f2fs_quota_write,
1889         .get_dquots     = f2fs_get_dquots,
1890 #endif
1891         .evict_inode    = f2fs_evict_inode,
1892         .put_super      = f2fs_put_super,
1893         .sync_fs        = f2fs_sync_fs,
1894         .freeze_fs      = f2fs_freeze,
1895         .unfreeze_fs    = f2fs_unfreeze,
1896         .statfs         = f2fs_statfs,
1897         .remount_fs     = f2fs_remount,
1898 };
1899
1900 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1901 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1902 {
1903         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1904                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1905                                 ctx, len, NULL);
1906 }
1907
1908 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1909                                                         void *fs_data)
1910 {
1911         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1912
1913         /*
1914          * Encrypting the root directory is not allowed because fsck
1915          * expects lost+found directory to exist and remain unencrypted
1916          * if LOST_FOUND feature is enabled.
1917          *
1918          */
1919         if (f2fs_sb_has_lost_found(sbi->sb) &&
1920                         inode->i_ino == F2FS_ROOT_INO(sbi))
1921                 return -EPERM;
1922
1923         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1924                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1925                                 ctx, len, fs_data, XATTR_CREATE);
1926 }
1927
1928 static bool f2fs_dummy_context(struct inode *inode)
1929 {
1930         return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
1931 }
1932
1933 static unsigned f2fs_max_namelen(struct inode *inode)
1934 {
1935         return S_ISLNK(inode->i_mode) ?
1936                         inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1937 }
1938
1939 static const struct fscrypt_operations f2fs_cryptops = {
1940         .key_prefix     = "f2fs:",
1941         .get_context    = f2fs_get_context,
1942         .set_context    = f2fs_set_context,
1943         .dummy_context  = f2fs_dummy_context,
1944         .empty_dir      = f2fs_empty_dir,
1945         .max_namelen    = f2fs_max_namelen,
1946 };
1947 #endif
1948
1949 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1950                 u64 ino, u32 generation)
1951 {
1952         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1953         struct inode *inode;
1954
1955         if (check_nid_range(sbi, ino))
1956                 return ERR_PTR(-ESTALE);
1957
1958         /*
1959          * f2fs_iget isn't quite right if the inode is currently unallocated!
1960          * However f2fs_iget currently does appropriate checks to handle stale
1961          * inodes so everything is OK.
1962          */
1963         inode = f2fs_iget(sb, ino);
1964         if (IS_ERR(inode))
1965                 return ERR_CAST(inode);
1966         if (unlikely(generation && inode->i_generation != generation)) {
1967                 /* we didn't find the right inode.. */
1968                 iput(inode);
1969                 return ERR_PTR(-ESTALE);
1970         }
1971         return inode;
1972 }
1973
1974 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1975                 int fh_len, int fh_type)
1976 {
1977         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1978                                     f2fs_nfs_get_inode);
1979 }
1980
1981 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1982                 int fh_len, int fh_type)
1983 {
1984         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1985                                     f2fs_nfs_get_inode);
1986 }
1987
1988 static const struct export_operations f2fs_export_ops = {
1989         .fh_to_dentry = f2fs_fh_to_dentry,
1990         .fh_to_parent = f2fs_fh_to_parent,
1991         .get_parent = f2fs_get_parent,
1992 };
1993
1994 static loff_t max_file_blocks(void)
1995 {
1996         loff_t result = 0;
1997         loff_t leaf_count = ADDRS_PER_BLOCK;
1998
1999         /*
2000          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2001          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2002          * space in inode.i_addr, it will be more safe to reassign
2003          * result as zero.
2004          */
2005
2006         /* two direct node blocks */
2007         result += (leaf_count * 2);
2008
2009         /* two indirect node blocks */
2010         leaf_count *= NIDS_PER_BLOCK;
2011         result += (leaf_count * 2);
2012
2013         /* one double indirect node block */
2014         leaf_count *= NIDS_PER_BLOCK;
2015         result += leaf_count;
2016
2017         return result;
2018 }
2019
2020 static int __f2fs_commit_super(struct buffer_head *bh,
2021                         struct f2fs_super_block *super)
2022 {
2023         lock_buffer(bh);
2024         if (super)
2025                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2026         set_buffer_dirty(bh);
2027         unlock_buffer(bh);
2028
2029         /* it's rare case, we can do fua all the time */
2030         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2031 }
2032
2033 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2034                                         struct buffer_head *bh)
2035 {
2036         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2037                                         (bh->b_data + F2FS_SUPER_OFFSET);
2038         struct super_block *sb = sbi->sb;
2039         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2040         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2041         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2042         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2043         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2044         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2045         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2046         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2047         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2048         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2049         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2050         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2051         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2052         u64 main_end_blkaddr = main_blkaddr +
2053                                 (segment_count_main << log_blocks_per_seg);
2054         u64 seg_end_blkaddr = segment0_blkaddr +
2055                                 (segment_count << log_blocks_per_seg);
2056
2057         if (segment0_blkaddr != cp_blkaddr) {
2058                 f2fs_msg(sb, KERN_INFO,
2059                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2060                         segment0_blkaddr, cp_blkaddr);
2061                 return true;
2062         }
2063
2064         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2065                                                         sit_blkaddr) {
2066                 f2fs_msg(sb, KERN_INFO,
2067                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2068                         cp_blkaddr, sit_blkaddr,
2069                         segment_count_ckpt << log_blocks_per_seg);
2070                 return true;
2071         }
2072
2073         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2074                                                         nat_blkaddr) {
2075                 f2fs_msg(sb, KERN_INFO,
2076                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2077                         sit_blkaddr, nat_blkaddr,
2078                         segment_count_sit << log_blocks_per_seg);
2079                 return true;
2080         }
2081
2082         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2083                                                         ssa_blkaddr) {
2084                 f2fs_msg(sb, KERN_INFO,
2085                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2086                         nat_blkaddr, ssa_blkaddr,
2087                         segment_count_nat << log_blocks_per_seg);
2088                 return true;
2089         }
2090
2091         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2092                                                         main_blkaddr) {
2093                 f2fs_msg(sb, KERN_INFO,
2094                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2095                         ssa_blkaddr, main_blkaddr,
2096                         segment_count_ssa << log_blocks_per_seg);
2097                 return true;
2098         }
2099
2100         if (main_end_blkaddr > seg_end_blkaddr) {
2101                 f2fs_msg(sb, KERN_INFO,
2102                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2103                         main_blkaddr,
2104                         segment0_blkaddr +
2105                                 (segment_count << log_blocks_per_seg),
2106                         segment_count_main << log_blocks_per_seg);
2107                 return true;
2108         } else if (main_end_blkaddr < seg_end_blkaddr) {
2109                 int err = 0;
2110                 char *res;
2111
2112                 /* fix in-memory information all the time */
2113                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2114                                 segment0_blkaddr) >> log_blocks_per_seg);
2115
2116                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2117                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2118                         res = "internally";
2119                 } else {
2120                         err = __f2fs_commit_super(bh, NULL);
2121                         res = err ? "failed" : "done";
2122                 }
2123                 f2fs_msg(sb, KERN_INFO,
2124                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
2125                         res, main_blkaddr,
2126                         segment0_blkaddr +
2127                                 (segment_count << log_blocks_per_seg),
2128                         segment_count_main << log_blocks_per_seg);
2129                 if (err)
2130                         return true;
2131         }
2132         return false;
2133 }
2134
2135 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2136                                 struct buffer_head *bh)
2137 {
2138         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2139                                         (bh->b_data + F2FS_SUPER_OFFSET);
2140         struct super_block *sb = sbi->sb;
2141         unsigned int blocksize;
2142
2143         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2144                 f2fs_msg(sb, KERN_INFO,
2145                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
2146                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2147                 return 1;
2148         }
2149
2150         /* Currently, support only 4KB page cache size */
2151         if (F2FS_BLKSIZE != PAGE_SIZE) {
2152                 f2fs_msg(sb, KERN_INFO,
2153                         "Invalid page_cache_size (%lu), supports only 4KB\n",
2154                         PAGE_SIZE);
2155                 return 1;
2156         }
2157
2158         /* Currently, support only 4KB block size */
2159         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2160         if (blocksize != F2FS_BLKSIZE) {
2161                 f2fs_msg(sb, KERN_INFO,
2162                         "Invalid blocksize (%u), supports only 4KB\n",
2163                         blocksize);
2164                 return 1;
2165         }
2166
2167         /* check log blocks per segment */
2168         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2169                 f2fs_msg(sb, KERN_INFO,
2170                         "Invalid log blocks per segment (%u)\n",
2171                         le32_to_cpu(raw_super->log_blocks_per_seg));
2172                 return 1;
2173         }
2174
2175         /* Currently, support 512/1024/2048/4096 bytes sector size */
2176         if (le32_to_cpu(raw_super->log_sectorsize) >
2177                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2178                 le32_to_cpu(raw_super->log_sectorsize) <
2179                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2180                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2181                         le32_to_cpu(raw_super->log_sectorsize));
2182                 return 1;
2183         }
2184         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2185                 le32_to_cpu(raw_super->log_sectorsize) !=
2186                         F2FS_MAX_LOG_SECTOR_SIZE) {
2187                 f2fs_msg(sb, KERN_INFO,
2188                         "Invalid log sectors per block(%u) log sectorsize(%u)",
2189                         le32_to_cpu(raw_super->log_sectors_per_block),
2190                         le32_to_cpu(raw_super->log_sectorsize));
2191                 return 1;
2192         }
2193
2194         /* check reserved ino info */
2195         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2196                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2197                 le32_to_cpu(raw_super->root_ino) != 3) {
2198                 f2fs_msg(sb, KERN_INFO,
2199                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2200                         le32_to_cpu(raw_super->node_ino),
2201                         le32_to_cpu(raw_super->meta_ino),
2202                         le32_to_cpu(raw_super->root_ino));
2203                 return 1;
2204         }
2205
2206         if (le32_to_cpu(raw_super->segment_count) > F2FS_MAX_SEGMENT) {
2207                 f2fs_msg(sb, KERN_INFO,
2208                         "Invalid segment count (%u)",
2209                         le32_to_cpu(raw_super->segment_count));
2210                 return 1;
2211         }
2212
2213         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2214         if (sanity_check_area_boundary(sbi, bh))
2215                 return 1;
2216
2217         return 0;
2218 }
2219
2220 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
2221 {
2222         unsigned int total, fsmeta;
2223         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2224         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2225         unsigned int ovp_segments, reserved_segments;
2226         unsigned int main_segs, blocks_per_seg;
2227         int i;
2228
2229         total = le32_to_cpu(raw_super->segment_count);
2230         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2231         fsmeta += le32_to_cpu(raw_super->segment_count_sit);
2232         fsmeta += le32_to_cpu(raw_super->segment_count_nat);
2233         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2234         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2235
2236         if (unlikely(fsmeta >= total))
2237                 return 1;
2238
2239         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2240         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2241
2242         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2243                         ovp_segments == 0 || reserved_segments == 0)) {
2244                 f2fs_msg(sbi->sb, KERN_ERR,
2245                         "Wrong layout: check mkfs.f2fs version");
2246                 return 1;
2247         }
2248
2249         main_segs = le32_to_cpu(raw_super->segment_count_main);
2250         blocks_per_seg = sbi->blocks_per_seg;
2251
2252         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2253                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2254                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2255                         return 1;
2256         }
2257         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2258                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2259                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2260                         return 1;
2261         }
2262
2263         if (unlikely(f2fs_cp_error(sbi))) {
2264                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2265                 return 1;
2266         }
2267         return 0;
2268 }
2269
2270 static void init_sb_info(struct f2fs_sb_info *sbi)
2271 {
2272         struct f2fs_super_block *raw_super = sbi->raw_super;
2273         int i, j;
2274
2275         sbi->log_sectors_per_block =
2276                 le32_to_cpu(raw_super->log_sectors_per_block);
2277         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2278         sbi->blocksize = 1 << sbi->log_blocksize;
2279         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2280         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2281         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2282         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2283         sbi->total_sections = le32_to_cpu(raw_super->section_count);
2284         sbi->total_node_count =
2285                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2286                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2287         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2288         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2289         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2290         sbi->cur_victim_sec = NULL_SECNO;
2291         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2292
2293         sbi->dir_level = DEF_DIR_LEVEL;
2294         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2295         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2296         clear_sbi_flag(sbi, SBI_NEED_FSCK);
2297
2298         for (i = 0; i < NR_COUNT_TYPE; i++)
2299                 atomic_set(&sbi->nr_pages[i], 0);
2300
2301         atomic_set(&sbi->wb_sync_req, 0);
2302
2303         INIT_LIST_HEAD(&sbi->s_list);
2304         mutex_init(&sbi->umount_mutex);
2305         for (i = 0; i < NR_PAGE_TYPE - 1; i++)
2306                 for (j = HOT; j < NR_TEMP_TYPE; j++)
2307                         mutex_init(&sbi->wio_mutex[i][j]);
2308         spin_lock_init(&sbi->cp_lock);
2309
2310         sbi->dirty_device = 0;
2311         spin_lock_init(&sbi->dev_lock);
2312
2313         init_rwsem(&sbi->sb_lock);
2314 }
2315
2316 static int init_percpu_info(struct f2fs_sb_info *sbi)
2317 {
2318         int err;
2319
2320         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2321         if (err)
2322                 return err;
2323
2324         return percpu_counter_init(&sbi->total_valid_inode_count, 0,
2325                                                                 GFP_KERNEL);
2326 }
2327
2328 #ifdef CONFIG_BLK_DEV_ZONED
2329 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2330 {
2331         struct block_device *bdev = FDEV(devi).bdev;
2332         sector_t nr_sectors = bdev->bd_part->nr_sects;
2333         sector_t sector = 0;
2334         struct blk_zone *zones;
2335         unsigned int i, nr_zones;
2336         unsigned int n = 0;
2337         int err = -EIO;
2338
2339         if (!f2fs_sb_has_blkzoned(sbi->sb))
2340                 return 0;
2341
2342         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2343                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2344                 return -EINVAL;
2345         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2346         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2347                                 __ilog2_u32(sbi->blocks_per_blkz))
2348                 return -EINVAL;
2349         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2350         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2351                                         sbi->log_blocks_per_blkz;
2352         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2353                 FDEV(devi).nr_blkz++;
2354
2355         FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2356                                                                 GFP_KERNEL);
2357         if (!FDEV(devi).blkz_type)
2358                 return -ENOMEM;
2359
2360 #define F2FS_REPORT_NR_ZONES   4096
2361
2362         zones = f2fs_kzalloc(sbi, sizeof(struct blk_zone) *
2363                                 F2FS_REPORT_NR_ZONES, GFP_KERNEL);
2364         if (!zones)
2365                 return -ENOMEM;
2366
2367         /* Get block zones type */
2368         while (zones && sector < nr_sectors) {
2369
2370                 nr_zones = F2FS_REPORT_NR_ZONES;
2371                 err = blkdev_report_zones(bdev, sector,
2372                                           zones, &nr_zones,
2373                                           GFP_KERNEL);
2374                 if (err)
2375                         break;
2376                 if (!nr_zones) {
2377                         err = -EIO;
2378                         break;
2379                 }
2380
2381                 for (i = 0; i < nr_zones; i++) {
2382                         FDEV(devi).blkz_type[n] = zones[i].type;
2383                         sector += zones[i].len;
2384                         n++;
2385                 }
2386         }
2387
2388         kfree(zones);
2389
2390         return err;
2391 }
2392 #endif
2393
2394 /*
2395  * Read f2fs raw super block.
2396  * Because we have two copies of super block, so read both of them
2397  * to get the first valid one. If any one of them is broken, we pass
2398  * them recovery flag back to the caller.
2399  */
2400 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2401                         struct f2fs_super_block **raw_super,
2402                         int *valid_super_block, int *recovery)
2403 {
2404         struct super_block *sb = sbi->sb;
2405         int block;
2406         struct buffer_head *bh;
2407         struct f2fs_super_block *super;
2408         int err = 0;
2409
2410         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2411         if (!super)
2412                 return -ENOMEM;
2413
2414         for (block = 0; block < 2; block++) {
2415                 bh = sb_bread(sb, block);
2416                 if (!bh) {
2417                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2418                                 block + 1);
2419                         err = -EIO;
2420                         continue;
2421                 }
2422
2423                 /* sanity checking of raw super */
2424                 if (sanity_check_raw_super(sbi, bh)) {
2425                         f2fs_msg(sb, KERN_ERR,
2426                                 "Can't find valid F2FS filesystem in %dth superblock",
2427                                 block + 1);
2428                         err = -EINVAL;
2429                         brelse(bh);
2430                         continue;
2431                 }
2432
2433                 if (!*raw_super) {
2434                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2435                                                         sizeof(*super));
2436                         *valid_super_block = block;
2437                         *raw_super = super;
2438                 }
2439                 brelse(bh);
2440         }
2441
2442         /* Fail to read any one of the superblocks*/
2443         if (err < 0)
2444                 *recovery = 1;
2445
2446         /* No valid superblock */
2447         if (!*raw_super)
2448                 kfree(super);
2449         else
2450                 err = 0;
2451
2452         return err;
2453 }
2454
2455 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2456 {
2457         struct buffer_head *bh;
2458         int err;
2459
2460         if ((recover && f2fs_readonly(sbi->sb)) ||
2461                                 bdev_read_only(sbi->sb->s_bdev)) {
2462                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2463                 return -EROFS;
2464         }
2465
2466         /* write back-up superblock first */
2467         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2468         if (!bh)
2469                 return -EIO;
2470         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2471         brelse(bh);
2472
2473         /* if we are in recovery path, skip writing valid superblock */
2474         if (recover || err)
2475                 return err;
2476
2477         /* write current valid superblock */
2478         bh = sb_bread(sbi->sb, sbi->valid_super_block);
2479         if (!bh)
2480                 return -EIO;
2481         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2482         brelse(bh);
2483         return err;
2484 }
2485
2486 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2487 {
2488         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2489         unsigned int max_devices = MAX_DEVICES;
2490         int i;
2491
2492         /* Initialize single device information */
2493         if (!RDEV(0).path[0]) {
2494                 if (!bdev_is_zoned(sbi->sb->s_bdev))
2495                         return 0;
2496                 max_devices = 1;
2497         }
2498
2499         /*
2500          * Initialize multiple devices information, or single
2501          * zoned block device information.
2502          */
2503         sbi->devs = f2fs_kzalloc(sbi, sizeof(struct f2fs_dev_info) *
2504                                                 max_devices, GFP_KERNEL);
2505         if (!sbi->devs)
2506                 return -ENOMEM;
2507
2508         for (i = 0; i < max_devices; i++) {
2509
2510                 if (i > 0 && !RDEV(i).path[0])
2511                         break;
2512
2513                 if (max_devices == 1) {
2514                         /* Single zoned block device mount */
2515                         FDEV(0).bdev =
2516                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2517                                         sbi->sb->s_mode, sbi->sb->s_type);
2518                 } else {
2519                         /* Multi-device mount */
2520                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2521                         FDEV(i).total_segments =
2522                                 le32_to_cpu(RDEV(i).total_segments);
2523                         if (i == 0) {
2524                                 FDEV(i).start_blk = 0;
2525                                 FDEV(i).end_blk = FDEV(i).start_blk +
2526                                     (FDEV(i).total_segments <<
2527                                     sbi->log_blocks_per_seg) - 1 +
2528                                     le32_to_cpu(raw_super->segment0_blkaddr);
2529                         } else {
2530                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2531                                 FDEV(i).end_blk = FDEV(i).start_blk +
2532                                         (FDEV(i).total_segments <<
2533                                         sbi->log_blocks_per_seg) - 1;
2534                         }
2535                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2536                                         sbi->sb->s_mode, sbi->sb->s_type);
2537                 }
2538                 if (IS_ERR(FDEV(i).bdev))
2539                         return PTR_ERR(FDEV(i).bdev);
2540
2541                 /* to release errored devices */
2542                 sbi->s_ndevs = i + 1;
2543
2544 #ifdef CONFIG_BLK_DEV_ZONED
2545                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
2546                                 !f2fs_sb_has_blkzoned(sbi->sb)) {
2547                         f2fs_msg(sbi->sb, KERN_ERR,
2548                                 "Zoned block device feature not enabled\n");
2549                         return -EINVAL;
2550                 }
2551                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
2552                         if (init_blkz_info(sbi, i)) {
2553                                 f2fs_msg(sbi->sb, KERN_ERR,
2554                                         "Failed to initialize F2FS blkzone information");
2555                                 return -EINVAL;
2556                         }
2557                         if (max_devices == 1)
2558                                 break;
2559                         f2fs_msg(sbi->sb, KERN_INFO,
2560                                 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
2561                                 i, FDEV(i).path,
2562                                 FDEV(i).total_segments,
2563                                 FDEV(i).start_blk, FDEV(i).end_blk,
2564                                 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
2565                                 "Host-aware" : "Host-managed");
2566                         continue;
2567                 }
2568 #endif
2569                 f2fs_msg(sbi->sb, KERN_INFO,
2570                         "Mount Device [%2d]: %20s, %8u, %8x - %8x",
2571                                 i, FDEV(i).path,
2572                                 FDEV(i).total_segments,
2573                                 FDEV(i).start_blk, FDEV(i).end_blk);
2574         }
2575         f2fs_msg(sbi->sb, KERN_INFO,
2576                         "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
2577         return 0;
2578 }
2579
2580 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
2581 {
2582         struct f2fs_sm_info *sm_i = SM_I(sbi);
2583
2584         /* adjust parameters according to the volume size */
2585         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
2586                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2587                 sm_i->dcc_info->discard_granularity = 1;
2588                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
2589         }
2590 }
2591
2592 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
2593 {
2594         struct f2fs_sb_info *sbi;
2595         struct f2fs_super_block *raw_super;
2596         struct inode *root;
2597         int err;
2598         bool retry = true, need_fsck = false;
2599         char *options = NULL;
2600         int recovery, i, valid_super_block;
2601         struct curseg_info *seg_i;
2602
2603 try_onemore:
2604         err = -EINVAL;
2605         raw_super = NULL;
2606         valid_super_block = -1;
2607         recovery = 0;
2608
2609         /* allocate memory for f2fs-specific super block info */
2610         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
2611         if (!sbi)
2612                 return -ENOMEM;
2613
2614         sbi->sb = sb;
2615
2616         /* Load the checksum driver */
2617         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
2618         if (IS_ERR(sbi->s_chksum_driver)) {
2619                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
2620                 err = PTR_ERR(sbi->s_chksum_driver);
2621                 sbi->s_chksum_driver = NULL;
2622                 goto free_sbi;
2623         }
2624
2625         /* set a block size */
2626         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
2627                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
2628                 goto free_sbi;
2629         }
2630
2631         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
2632                                                                 &recovery);
2633         if (err)
2634                 goto free_sbi;
2635
2636         sb->s_fs_info = sbi;
2637         sbi->raw_super = raw_super;
2638
2639         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2640         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2641
2642         /* precompute checksum seed for metadata */
2643         if (f2fs_sb_has_inode_chksum(sb))
2644                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
2645                                                 sizeof(raw_super->uuid));
2646
2647         /*
2648          * The BLKZONED feature indicates that the drive was formatted with
2649          * zone alignment optimization. This is optional for host-aware
2650          * devices, but mandatory for host-managed zoned block devices.
2651          */
2652 #ifndef CONFIG_BLK_DEV_ZONED
2653         if (f2fs_sb_has_blkzoned(sb)) {
2654                 f2fs_msg(sb, KERN_ERR,
2655                          "Zoned block device support is not enabled\n");
2656                 err = -EOPNOTSUPP;
2657                 goto free_sb_buf;
2658         }
2659 #endif
2660         default_options(sbi);
2661         /* parse mount options */
2662         options = kstrdup((const char *)data, GFP_KERNEL);
2663         if (data && !options) {
2664                 err = -ENOMEM;
2665                 goto free_sb_buf;
2666         }
2667
2668         err = parse_options(sb, options);
2669         if (err)
2670                 goto free_options;
2671
2672         sbi->max_file_blocks = max_file_blocks();
2673         sb->s_maxbytes = sbi->max_file_blocks <<
2674                                 le32_to_cpu(raw_super->log_blocksize);
2675         sb->s_max_links = F2FS_LINK_MAX;
2676         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2677
2678 #ifdef CONFIG_QUOTA
2679         sb->dq_op = &f2fs_quota_operations;
2680         if (f2fs_sb_has_quota_ino(sb))
2681                 sb->s_qcop = &dquot_quotactl_sysfile_ops;
2682         else
2683                 sb->s_qcop = &f2fs_quotactl_ops;
2684         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
2685
2686         if (f2fs_sb_has_quota_ino(sbi->sb)) {
2687                 for (i = 0; i < MAXQUOTAS; i++) {
2688                         if (f2fs_qf_ino(sbi->sb, i))
2689                                 sbi->nquota_files++;
2690                 }
2691         }
2692 #endif
2693
2694         sb->s_op = &f2fs_sops;
2695 #ifdef CONFIG_F2FS_FS_ENCRYPTION
2696         sb->s_cop = &f2fs_cryptops;
2697 #endif
2698         sb->s_xattr = f2fs_xattr_handlers;
2699         sb->s_export_op = &f2fs_export_ops;
2700         sb->s_magic = F2FS_SUPER_MAGIC;
2701         sb->s_time_gran = 1;
2702         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2703                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2704         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
2705         sb->s_iflags |= SB_I_CGROUPWB;
2706
2707         /* init f2fs-specific super block info */
2708         sbi->valid_super_block = valid_super_block;
2709         mutex_init(&sbi->gc_mutex);
2710         mutex_init(&sbi->cp_mutex);
2711         init_rwsem(&sbi->node_write);
2712         init_rwsem(&sbi->node_change);
2713
2714         /* disallow all the data/node/meta page writes */
2715         set_sbi_flag(sbi, SBI_POR_DOING);
2716         spin_lock_init(&sbi->stat_lock);
2717
2718         /* init iostat info */
2719         spin_lock_init(&sbi->iostat_lock);
2720         sbi->iostat_enable = false;
2721
2722         for (i = 0; i < NR_PAGE_TYPE; i++) {
2723                 int n = (i == META) ? 1: NR_TEMP_TYPE;
2724                 int j;
2725
2726                 sbi->write_io[i] = f2fs_kmalloc(sbi,
2727                                         n * sizeof(struct f2fs_bio_info),
2728                                         GFP_KERNEL);
2729                 if (!sbi->write_io[i]) {
2730                         err = -ENOMEM;
2731                         goto free_options;
2732                 }
2733
2734                 for (j = HOT; j < n; j++) {
2735                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
2736                         sbi->write_io[i][j].sbi = sbi;
2737                         sbi->write_io[i][j].bio = NULL;
2738                         spin_lock_init(&sbi->write_io[i][j].io_lock);
2739                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
2740                 }
2741         }
2742
2743         init_rwsem(&sbi->cp_rwsem);
2744         init_waitqueue_head(&sbi->cp_wait);
2745         init_sb_info(sbi);
2746
2747         err = init_percpu_info(sbi);
2748         if (err)
2749                 goto free_bio_info;
2750
2751         if (F2FS_IO_SIZE(sbi) > 1) {
2752                 sbi->write_io_dummy =
2753                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
2754                 if (!sbi->write_io_dummy) {
2755                         err = -ENOMEM;
2756                         goto free_percpu;
2757                 }
2758         }
2759
2760         /* get an inode for meta space */
2761         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
2762         if (IS_ERR(sbi->meta_inode)) {
2763                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
2764                 err = PTR_ERR(sbi->meta_inode);
2765                 goto free_io_dummy;
2766         }
2767
2768         err = get_valid_checkpoint(sbi);
2769         if (err) {
2770                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
2771                 goto free_meta_inode;
2772         }
2773
2774         /* Initialize device list */
2775         err = f2fs_scan_devices(sbi);
2776         if (err) {
2777                 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
2778                 goto free_devices;
2779         }
2780
2781         sbi->total_valid_node_count =
2782                                 le32_to_cpu(sbi->ckpt->valid_node_count);
2783         percpu_counter_set(&sbi->total_valid_inode_count,
2784                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
2785         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
2786         sbi->total_valid_block_count =
2787                                 le64_to_cpu(sbi->ckpt->valid_block_count);
2788         sbi->last_valid_block_count = sbi->total_valid_block_count;
2789         sbi->reserved_blocks = 0;
2790         sbi->current_reserved_blocks = 0;
2791         limit_reserve_root(sbi);
2792
2793         for (i = 0; i < NR_INODE_TYPE; i++) {
2794                 INIT_LIST_HEAD(&sbi->inode_list[i]);
2795                 spin_lock_init(&sbi->inode_lock[i]);
2796         }
2797
2798         init_extent_cache_info(sbi);
2799
2800         init_ino_entry_info(sbi);
2801
2802         /* setup f2fs internal modules */
2803         err = build_segment_manager(sbi);
2804         if (err) {
2805                 f2fs_msg(sb, KERN_ERR,
2806                         "Failed to initialize F2FS segment manager");
2807                 goto free_sm;
2808         }
2809         err = build_node_manager(sbi);
2810         if (err) {
2811                 f2fs_msg(sb, KERN_ERR,
2812                         "Failed to initialize F2FS node manager");
2813                 goto free_nm;
2814         }
2815
2816         /* For write statistics */
2817         if (sb->s_bdev->bd_part)
2818                 sbi->sectors_written_start =
2819                         (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2820
2821         /* Read accumulated write IO statistics if exists */
2822         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
2823         if (__exist_node_summaries(sbi))
2824                 sbi->kbytes_written =
2825                         le64_to_cpu(seg_i->journal->info.kbytes_written);
2826
2827         build_gc_manager(sbi);
2828
2829         /* get an inode for node space */
2830         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
2831         if (IS_ERR(sbi->node_inode)) {
2832                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
2833                 err = PTR_ERR(sbi->node_inode);
2834                 goto free_nm;
2835         }
2836
2837         err = f2fs_build_stats(sbi);
2838         if (err)
2839                 goto free_node_inode;
2840
2841         /* read root inode and dentry */
2842         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
2843         if (IS_ERR(root)) {
2844                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
2845                 err = PTR_ERR(root);
2846                 goto free_stats;
2847         }
2848         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2849                 iput(root);
2850                 err = -EINVAL;
2851                 goto free_node_inode;
2852         }
2853
2854         sb->s_root = d_make_root(root); /* allocate root dentry */
2855         if (!sb->s_root) {
2856                 err = -ENOMEM;
2857                 goto free_root_inode;
2858         }
2859
2860         err = f2fs_register_sysfs(sbi);
2861         if (err)
2862                 goto free_root_inode;
2863
2864 #ifdef CONFIG_QUOTA
2865         /*
2866          * Turn on quotas which were not enabled for read-only mounts if
2867          * filesystem has quota feature, so that they are updated correctly.
2868          */
2869         if (f2fs_sb_has_quota_ino(sb) && !f2fs_readonly(sb)) {
2870                 err = f2fs_enable_quotas(sb);
2871                 if (err) {
2872                         f2fs_msg(sb, KERN_ERR,
2873                                 "Cannot turn on quotas: error %d", err);
2874                         goto free_sysfs;
2875                 }
2876         }
2877 #endif
2878         /* if there are nt orphan nodes free them */
2879         err = recover_orphan_inodes(sbi);
2880         if (err)
2881                 goto free_meta;
2882
2883         /* recover fsynced data */
2884         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
2885                 /*
2886                  * mount should be failed, when device has readonly mode, and
2887                  * previous checkpoint was not done by clean system shutdown.
2888                  */
2889                 if (bdev_read_only(sb->s_bdev) &&
2890                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
2891                         err = -EROFS;
2892                         goto free_meta;
2893                 }
2894
2895                 if (need_fsck)
2896                         set_sbi_flag(sbi, SBI_NEED_FSCK);
2897
2898                 if (!retry)
2899                         goto skip_recovery;
2900
2901                 err = recover_fsync_data(sbi, false);
2902                 if (err < 0) {
2903                         need_fsck = true;
2904                         f2fs_msg(sb, KERN_ERR,
2905                                 "Cannot recover all fsync data errno=%d", err);
2906                         goto free_meta;
2907                 }
2908         } else {
2909                 err = recover_fsync_data(sbi, true);
2910
2911                 if (!f2fs_readonly(sb) && err > 0) {
2912                         err = -EINVAL;
2913                         f2fs_msg(sb, KERN_ERR,
2914                                 "Need to recover fsync data");
2915                         goto free_meta;
2916                 }
2917         }
2918 skip_recovery:
2919         /* recover_fsync_data() cleared this already */
2920         clear_sbi_flag(sbi, SBI_POR_DOING);
2921
2922         /*
2923          * If filesystem is not mounted as read-only then
2924          * do start the gc_thread.
2925          */
2926         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2927                 /* After POR, we can run background GC thread.*/
2928                 err = start_gc_thread(sbi);
2929                 if (err)
2930                         goto free_meta;
2931         }
2932         kfree(options);
2933
2934         /* recover broken superblock */
2935         if (recovery) {
2936                 err = f2fs_commit_super(sbi, true);
2937                 f2fs_msg(sb, KERN_INFO,
2938                         "Try to recover %dth superblock, ret: %d",
2939                         sbi->valid_super_block ? 1 : 2, err);
2940         }
2941
2942         f2fs_join_shrinker(sbi);
2943
2944         f2fs_tuning_parameters(sbi);
2945
2946         f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
2947                                 cur_cp_version(F2FS_CKPT(sbi)));
2948         f2fs_update_time(sbi, CP_TIME);
2949         f2fs_update_time(sbi, REQ_TIME);
2950         return 0;
2951
2952 free_meta:
2953 #ifdef CONFIG_QUOTA
2954         if (f2fs_sb_has_quota_ino(sb) && !f2fs_readonly(sb))
2955                 f2fs_quota_off_umount(sbi->sb);
2956 #endif
2957         f2fs_sync_inode_meta(sbi);
2958         /*
2959          * Some dirty meta pages can be produced by recover_orphan_inodes()
2960          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
2961          * followed by write_checkpoint() through f2fs_write_node_pages(), which
2962          * falls into an infinite loop in sync_meta_pages().
2963          */
2964         truncate_inode_pages_final(META_MAPPING(sbi));
2965 #ifdef CONFIG_QUOTA
2966 free_sysfs:
2967 #endif
2968         f2fs_unregister_sysfs(sbi);
2969 free_root_inode:
2970         dput(sb->s_root);
2971         sb->s_root = NULL;
2972 free_stats:
2973         f2fs_destroy_stats(sbi);
2974 free_node_inode:
2975         release_ino_entry(sbi, true);
2976         truncate_inode_pages_final(NODE_MAPPING(sbi));
2977         iput(sbi->node_inode);
2978 free_nm:
2979         destroy_node_manager(sbi);
2980 free_sm:
2981         destroy_segment_manager(sbi);
2982 free_devices:
2983         destroy_device_list(sbi);
2984         kfree(sbi->ckpt);
2985 free_meta_inode:
2986         make_bad_inode(sbi->meta_inode);
2987         iput(sbi->meta_inode);
2988 free_io_dummy:
2989         mempool_destroy(sbi->write_io_dummy);
2990 free_percpu:
2991         destroy_percpu_info(sbi);
2992 free_bio_info:
2993         for (i = 0; i < NR_PAGE_TYPE; i++)
2994                 kfree(sbi->write_io[i]);
2995 free_options:
2996 #ifdef CONFIG_QUOTA
2997         for (i = 0; i < MAXQUOTAS; i++)
2998                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2999 #endif
3000         kfree(options);
3001 free_sb_buf:
3002         kfree(raw_super);
3003 free_sbi:
3004         if (sbi->s_chksum_driver)
3005                 crypto_free_shash(sbi->s_chksum_driver);
3006         kfree(sbi);
3007
3008         /* give only one another chance */
3009         if (retry) {
3010                 retry = false;
3011                 shrink_dcache_sb(sb);
3012                 goto try_onemore;
3013         }
3014         return err;
3015 }
3016
3017 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3018                         const char *dev_name, void *data)
3019 {
3020         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3021 }
3022
3023 static void kill_f2fs_super(struct super_block *sb)
3024 {
3025         if (sb->s_root) {
3026                 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
3027                 stop_gc_thread(F2FS_SB(sb));
3028                 stop_discard_thread(F2FS_SB(sb));
3029         }
3030         kill_block_super(sb);
3031 }
3032
3033 static struct file_system_type f2fs_fs_type = {
3034         .owner          = THIS_MODULE,
3035         .name           = "f2fs",
3036         .mount          = f2fs_mount,
3037         .kill_sb        = kill_f2fs_super,
3038         .fs_flags       = FS_REQUIRES_DEV,
3039 };
3040 MODULE_ALIAS_FS("f2fs");
3041
3042 static int __init init_inodecache(void)
3043 {
3044         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3045                         sizeof(struct f2fs_inode_info), 0,
3046                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3047         if (!f2fs_inode_cachep)
3048                 return -ENOMEM;
3049         return 0;
3050 }
3051
3052 static void destroy_inodecache(void)
3053 {
3054         /*
3055          * Make sure all delayed rcu free inodes are flushed before we
3056          * destroy cache.
3057          */
3058         rcu_barrier();
3059         kmem_cache_destroy(f2fs_inode_cachep);
3060 }
3061
3062 static int __init init_f2fs_fs(void)
3063 {
3064         int err;
3065
3066         f2fs_build_trace_ios();
3067
3068         err = init_inodecache();
3069         if (err)
3070                 goto fail;
3071         err = create_node_manager_caches();
3072         if (err)
3073                 goto free_inodecache;
3074         err = create_segment_manager_caches();
3075         if (err)
3076                 goto free_node_manager_caches;
3077         err = create_checkpoint_caches();
3078         if (err)
3079                 goto free_segment_manager_caches;
3080         err = create_extent_cache();
3081         if (err)
3082                 goto free_checkpoint_caches;
3083         err = f2fs_init_sysfs();
3084         if (err)
3085                 goto free_extent_cache;
3086         err = register_shrinker(&f2fs_shrinker_info);
3087         if (err)
3088                 goto free_sysfs;
3089         err = register_filesystem(&f2fs_fs_type);
3090         if (err)
3091                 goto free_shrinker;
3092         err = f2fs_create_root_stats();
3093         if (err)
3094                 goto free_filesystem;
3095         return 0;
3096
3097 free_filesystem:
3098         unregister_filesystem(&f2fs_fs_type);
3099 free_shrinker:
3100         unregister_shrinker(&f2fs_shrinker_info);
3101 free_sysfs:
3102         f2fs_exit_sysfs();
3103 free_extent_cache:
3104         destroy_extent_cache();
3105 free_checkpoint_caches:
3106         destroy_checkpoint_caches();
3107 free_segment_manager_caches:
3108         destroy_segment_manager_caches();
3109 free_node_manager_caches:
3110         destroy_node_manager_caches();
3111 free_inodecache:
3112         destroy_inodecache();
3113 fail:
3114         return err;
3115 }
3116
3117 static void __exit exit_f2fs_fs(void)
3118 {
3119         f2fs_destroy_root_stats();
3120         unregister_filesystem(&f2fs_fs_type);
3121         unregister_shrinker(&f2fs_shrinker_info);
3122         f2fs_exit_sysfs();
3123         destroy_extent_cache();
3124         destroy_checkpoint_caches();
3125         destroy_segment_manager_caches();
3126         destroy_node_manager_caches();
3127         destroy_inodecache();
3128         f2fs_destroy_trace_ios();
3129 }
3130
3131 module_init(init_f2fs_fs)
3132 module_exit(exit_f2fs_fs)
3133
3134 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3135 MODULE_DESCRIPTION("Flash Friendly File System");
3136 MODULE_LICENSE("GPL");
3137